Linked Research on the Decentralised Web

In research projects, there are various stakeholders (in)directly acting as internal and external forces in scholarly communication. In order to address societal concerns, research sponsors fund projects to attain reusable technical designs and acquire new knowledge in the process. Knowledge workers like researchers and scholars are driven by curiosity to advance our understanding of the universe and share their findings. The public is interested in accessing applications of research to educate and for the well-being of life. The industry needs efficient functioning of scientific communication in order to translate the return of research capital.

I am a member of stakeholders acting in particular as a knowledge worker, a part of public, an operator (end user, maintainer, operational supporter).

There are several sociotechnical limitations in the contemporary paradigm that the scholarly communication operates in. I will summarise a subset of problem areas pertaining to interoperability of systems and autonomy of actors. Each problem space will be reviewed in relevant sections of this thesis.

There is an abundance of scientific information on the Web. However, humans and machines have restricted access, are required to pay, or hindered to efficiently discover, make sense of, and create new connections between information.

Data on the Web tends to be locked into non-interoperable Web systems or be only usable by applications which created it. Actors – creators and readers – in these systems often do not have the means to switch between applications and share data between them.

There is a lack of diverse value chains for different units of scholarly communication due to tightly coupled proprietary systems. The actors in these systems create and exchange information by (being forced into) using packaged systems and workflows, give up their rights, and not have control over how their creations can be disseminated and reused.

The actors in the system commonly use print-centric tools and data representations to exchange content on the Web and neglect to take advantage of the available affordances of the Web (media). Actors’ possibility to shape their contributions in Web media is typically constrained to third-party service provider’s centralised and non-interoperable solutions.

Actors’ use of personal and professional identifiers, as well as data storage from third-parties are subject to privacy issues. Actors typically manage multiple disconnected profiles, social connections, and data spaces.

These class of problems in the scholarly ecosystem have implications on access, interoperability, findability, cost and privacy. Consequently, resulting in global deficiencies in the ecosystem. The research questions; Technical Research Problems and Knowledge Questions, are aimed at understanding the internal mechanisms of these type of problems and address them for stakeholders.

Given the Problem Context, the technical goals of this thesis are to identify gaps in the Web standards landscape and to create new technical standards where necessary. A further goal is to demonstrate how existing and new standards and technologies can be used in combination as part of a user-facing application for publishing and interacting with scholarly artifacts.

The Research Questions are geared to meeting these goals.

The knowledge goals of this thesis describe the theoretical underpinnings necessary to accomplish the Technical Research Goals.

We need to understand the various technological mediums in which scholarly communication has historically operated, and how and why this has changed, in order to have a frame of reference for new work. Our investigation must examine scholarship with respect to the media employed for publication, dissemination and feedback, and identify some of the constraints and limitations. This research can serve to describe the effects and artifacts of scholarly communication, as well as conceptual and technical designs that are used to exchange units of information by both humans and machines.

The scholarly communication ecosystem is a complex sociotechnical system, which cannot be understood from either a social or a technical perspective alone. Understanding how these aspects feed into and influence on another is crucial for the furtherance of this work.

The corpus of knowledge collected in this thesis is driven by understanding the current state of knowledge in the field based on scientific, technical, and professional literature. Literature Review and Citations describes the constraints on how research knowledge will be collected.

The technical research problems here consist of artifacts to (re)design, requirements to satisfy, stakeholders’ goals to achieve, and the problem context. The overarching technical research problem is defined as follows:

Central Design Problem

How can we design decentralised Web applications for information exchange in an open and decoupled scholarly communication ecosystem?

This research problem can be refined into two sub-problems:

Mechanisms

What technical mechanisms, standards or protocols are necessary for decentralised information exchange on the Web? Which already exist and what is missing?

Artifacts

How can Web technologies be employed to fulfill the core functions of scholarly communication in an open and interoperable way?

The Knowledge Goals can be refined into following knowledge questions:

• How can we situate scholarly communication using the Web with respect to technological communication mediums and scientific paradigms?
• What are the core components of scientific communication, how are they configured, and what are the effects of the interactions between them?
• What are contemporary practices around knowledge creation and dissemination, and how are communities working to improve these?

This thesis is sectioned into following main topic areas:

Scholarly Communication on the Web

Answers knowledge questions to describe and explain characteristics of the Web and the state of scholarly communication on the Web.

Structure of Scholarly Information

Answers knowledge questions to identify common structural and semantic patterns for human- and machine-readable information. Contributions: Linked Statistics on publishing and exchanging statistical Linked Data.

Decentralising Scholarly Communication

Answers knowledge questions pertaining to design of socially-aware decentralised systems, as well as classification of Web specifications for forces and functions in scientific communication. Contributions: Degree of Control, Forces and Functions in Specifications, The Effects and Artifacts of Autonomous Engagement.

Linked Data Notifications

Answers technical research problems on how to design a decentralised notification protocol. Contributions: Design of A Decentralised Notifications Protocol to exchange reusable decentralised notifications.

Decentralised Linked Research Application

Answers technical research problems on how to design a decentralised application. Contributions: Linking the Decentralised Information Space, Implementing a Read-Write Application, Forces and Functions in dokieli.

Linked Research

A description and explanation of the effects of applying Web-centric standards and practices in order to enable creators and consumers of open knowledge, as well as a generalisation of core principles that can serve as an exemplar to set forth a shift in scholarly communication. Contributions: Design Principles, Call for Linked Research, Forces and Functions of Linked Research, Linked Research as a Paradigm.

In Conclusions, I will discuss the thesis findings in Research Questions Review, offer my Interpretations, and share Perspectives for the future.

The five review characteristics; focus, goal, perspective, coverage, and audience, are borrowed from the summary at Practical Assessment, Research & Evaluation, Randolph, 2009, which is based on Organizing knowledge syntheses: A taxonomy of literature reviews, Cooper, 1988.

Focus: The core focus of the reviews and cited material was to establish a connection between the available theories and their underlying mechanisms, as well as the application of certain interventions to help identify and apply a practical need. Hence, majority of the analysis and synthesis focuses on practices or applications in the field.

Goal: My intention is to present this thesis in a way that is technically and meaningfully connected with the knowledge out there, as well as to connect my contributions. Hence, I would like any reader (human, machine, or other) in the future to be able to observe and verify the connections that I have put together as much as possible. The main interest for the reviews and citations were information integration for the big picture, identification and discussion of central issues, and explicating arguments.

Perspective: As this thesis is about connecting and endorsing open scholarly knowledge in context of decentralised and social Web, there is a selection bias in favour of (research) information based on the following criteria:

• Publicly accessible URL in the case of digital objects.
• Referencable URI in the case of printed material.
• Article and data including full text and media.
• Accessible by anyone (humans and machines).
• Available to anyone without financial, legal, or technical barriers (besides an Internet connection, free of charge).
• Preference for source materials (as opposed to third-party re-publications).
• Preference for research artifacts that can be publicly archived and recalled with free on-demand services.
• Preference for research artifacts made available with declarative languages (as opposed to dynamically generated as part of an application).

At the time of this writing, HTTP URIs of research objects resolving to paywall or access toll landing pages were excluded. The rationale is that if information is only available to a subset of humans, then it is not deemed to be part of the observable and reproducible universal knowledge – a fundamental requirement of applying the scientific method. Fortunately, I was able to gather plenty of open and free resources to support the narrative of this thesis, and addressing its knowledge goals.

Coverage: The reviewed and cited works are composed of a purposive sample, examining only pivotal works and the source of concepts in the field to tie them together. While this selection is naturally not exhaustive in and itself, I have considered, located, and discussed material that was available at the time of my investigation. While there may be natural imperfections in data collection, evaluation, analysis, and interpretation in any literature review and bibliographic citations, I have aimed to be systematic in what’s covered. To that end, the works – text, data, media, or other – mentioned in this thesis were included or excluded based on the following criteria:

• Information in English.
• Web accessible digital objects.
• Printed material available through a public library loan.
• Referencing information that is legally made available to the public (to the best of my knowledge!)

Organisation: Historical and conceptual formats were used as organisation schemes. My observations and categorisations are in Scholarly Communication on the Web, Structure of Scholarly Information and Decentralising Scholarly Communication are first organised conceptually, and then chronologically ordered.

Audience: for anyone interested in traversing the connections and building on the underlying findings. See also Audience.

Unless otherwise noted, all of the quotations are sic.

With the exception of referring to existing literature, references made to research, scientific or scholarly communication in this thesis is discipline agnostic. Hence, I mean any academic discipline, from humanities, social, natural, formal, to applied sciences.

The prefixes and namespaces that are used in this article are as follows:

acl

http://www.w3.org/ns/auth/acl#

as

https://www.w3.org/ns/activitystreams#

cert

http://www.w3.org/ns/auth/cert#

cito

http://purl.org/spar/cito/

contact

http://www.w3.org/2000/10/swap/pim/contact#

dcterms

http://purl.org/dc/terms/

doap

http://usefulinc.com/ns/doap#

earl

http://www.w3.org/ns/earl#

foaf

http://xmlns.com/foaf/0.1/

ldn

https://www.w3.org/TR/ldn/#

ldp

http://www.w3.org/ns/ldp#

mem

http://mementoweb.org/ns#

oa

http://www.w3.org/ns/oa#

owl

http://www.w3.org/2002/07/owl#

pim

http://www.w3.org/ns/pim/space#

prov

http://www.w3.org/ns/prov#

qb

http://purl.org/linked-data/cube#

rdf

http://www.w3.org/1999/02/22-rdf-syntax-ns#

rdfs

http://www.w3.org/2000/01/rdf-schema#

solid

http://www.w3.org/ns/solid/terms#

schema

http://schema.org/

skos

http://www.w3.org/2004/02/skos/core#

void

http://rdfs.org/ns/void#

xhv

http://www.w3.org/1999/xhtml/vocab#

This thesis should be read like all other theses. First, it should be read cover-to-cover, multiple times. Then, it should be read backwards at least once. Then it should be read by picking random sections from the contents list and following all the cross-references.

Gutenberg had, in effect, made every man a reader. Today, Xerox and other forms of reprography tend to make every man a publisher. This massive reversal has, for one of its consequences, elitism. The nature of the mass production of uniform volumes certainly did not foster elites but rather habits of universal reading. Paradoxically, when there are many readers, the author can wield great private power, whereas small reading elites may exert large corporate power.

In the 1962 book, The Gutenberg Galaxy, Marshall McLuhan explains how the adoption of new technological mediums shifted human culture in Europe through four eras, with emphasis placed on extending our senses:

1. acoustic age (c. until 800 BCE); audile-tactile, preliterate tribal culture,
2. age of writing (c. 800 BCE to 1500 CE); audile/visual, manuscript culture,
3. age of print (c. 1500 to 1850 CE); visual, mechanical mass communication,
4. age of electronic media (c. 1850 CE to present); central nervous system, global village

Acoustic age: According to McLuhan, the acoustic world depended on speech and hearing for communication. Information was an instantaneous and simultaneous experience for speakers and listeners with a collective identity. Existence was about narrated events and survived through story-telling, songs, and poems.

Writing age: When the phonetic alphabet came along, information could be defined, classified, and referenced, as well as better preserved. The written word made human knowledge tangible, stretching itself across time and space. While the written word increased the need of our visual system, it brought along the acoustic world with it – early reading was still an oral, and by and large an external (social) activity, as opposed to internal in one’s mind.

Print age: Through the age of writing to the age of print, the phonetic alphabet transformed from being an audile-tactile experience to purely visual. Gutenberg’s printing press (c. 1439 CE), through the process of mechanical mass production, had an extensive impact on the circulation of information, and triggered reconfiguration of societies’ structures. Books have boundaries, uniformity, continuity, linearity, and repeatability; this had a profound influence on the organisation of social systems. Print further detribalised social organisations, enabling individualism, fragmentation, government centralism, and nationalism. As a work was mechanically copied, it maintained its consistency and accuracy in comparison to hand copied manuscripts. It was easier to transmit information without personal contact. Then came the responsibility for authors to be consistent with their definitive statements. In Connections, 1978, Burke, posits that once a piece of knowledge was publicly shared, its creator was identifiable and received recognition. Print media democratised knowledge as soon as it became a commodity, mass produced and distributed to populations. The general public had a reason to not only access or acquire media like books, but also had a reason to learn to read for themselves, where previously it was necessary to go through a third-party; the wealthy, scholar, or theologian, in order to be informed. The increase in literacy consequently helped the public to be better knowledgeable about world matters without being subject to official or unofficial censorship. As there were more readers, the ability to write fostered the formulation and preservation of individual thought and culture, as well as enabling the public to voice themselves and question authority. In The Printing Press as an Agent of Change, Eisenstein, 1979, contends that the print media also played a key role in facilitating scientific publishing and the scientific revolution.

Electronic age: The electric mass-age made it possible to have instantaneous communication in different forms across the globe. Electronic media did not have the same kinds of boundaries that the print had on society and so society was able to re-organise itself differently. Unlike print media, electronic media like telephone, radio, and television, did not require literacy, and consequently the support of a stable education system. As a whole, it had the same effects on society as the acoustic space created in the tribal world. The speed of information increased, and with it recovered sociality and involvement which was previously abandoned as the primary mode of information exchange. Perhaps more interestingly, as the flow of information through various electronic media accelerated, it facilitated our natural tendency to recognise patterns and (ir)regularities in data – in comparison to print media which happened to isolate, classify, and immobilise items in fixed space.

Later I will revisit McLuhan’s studies on the psychic and social consequences of technological innovation and apply them to scholarly communication on the Web.

I simply assert the existence of significant limits to what the proponents of different theories can communicate to one another.

In The Structure of Scientific Revolutions, Thomas Kuhn, 1962, proposes an explanation for the process of discovery and progress in the history and philosophy of science. Kuhn contended that science did not progress linearly through history, rather it starts with a period that’s disorganised, and then a paradigm – a collection of exemplar agreed by scientists on how problems are to be understood – develops to organise things for day-to-day normal science to take place under the governed paradigm. After some time, anomalies in the paradigm accumulate and lead to a crisis phase where multiple competing paradigms attempt to resolve the issues. This is a particular point in which the community struggles to come to an agreement on the fittest way to practise future science. Finally a paradigm shift – revolution – refers to the last phase where transforming ideas and assertions mature enough to bring order to earlier anomalies, and cause a change in world view.

In Objectivity, Value Judgment, and Theory Choice, Kuhn, 1973, sets out five characteristics as a basis to explain which paradigms and methods have been adopted in the past, or may be considered for adoption in the future:

Accuracy

Application of quantitative and qualitative agreements in a domain.

Consistency

Internal and external consistency in relation to related theories.

Scope

Coverage and consequences of theories in what they can explain.

Simplicity

Computational labour required for explanations.

Fruitfulness

Explaining previously unnoted relationships or potential to disclose new phenomenon.

Kuhn reasoned that progress in science was not a mere line leading to truth, but the notion of moving away from less adequate concepts of and interactions with the world in favour of more fruitful. As to which paradigm can be objectively determined to be better, Kuhn argued that scientists’ decision was ultimately inter-subjective and required value judgement because paradigms – the ideas and assertions within – were not necessarily directly comparable. For example, early versions of the heliocentric model proposed by Copernicus in Commentariolus (1514), as well as the matured version in De revolutionibus orbium coelestium (1543) was neither quantitatively or qualitatively more accurate than Ptolemy’s geocentric model (c. 2nd century CE) based on the knowledge and instruments available at the time. While they were equally internally consistent, geocentrism was regarded to be more consistent with the other accepted theories and doctrines in comparison. Copernicus’ case however offered a simpler explanation and broader scope of application which ultimately was preferred. These were attractive characteristics for Kepler and Galileo when they came to investigate. With Newton’s laws for motion and gravitation, the heliocentric model matured enough to set a new course for science and research.

Another example demonstrating the complexity of choosing a theory was from the perspective of language and communication. For instance, the term "mass" has radically different meanings in Newton’s theory of gravity and Einstein’s relativity theory because the definitions are isolated to their respective paradigms. However, General Relativity gave a new lens to understanding the interaction of mass in space-time. While the definition for mass changed, Einstein’s theory was still able to precisely predict Newton’s theory, and further provided an explanation for how gravity worked.

Contemporary scholarly communication is situated within a particular paradigm. In Linked Research as a Paradigm I will use Kuhn’s theory of choosing paradigms as a heuristic device to Contextualise Linked Research.

The Gutenberg Galaxy and The Structure of Scientific Revolutions were coincidently published in 1962. While the accomplishments of McLuhan and Kuhn had differences in methodology and structure, they reveal the overwhelming effects of patterns of change when new technologies and paradigms are adopted by society. These major communication shifts in society as a whole, and specifically in the mode of scientific inquiry, are rare and significant events in history. Content follows form; mediums and paradigms shape the space.

As the characteristics of the spoken word carried over to writing, and writing to print, print media also influenced electronic media like the Web and hypermedia. Print’s typography and knowledge organisation affects the standards, tools, and interfaces that are still used today to exchange knowledge on the Web, as well as certain social expectations or assumptions. For example, authoring tools help us to design and create electronic documents that resemble print representations even if they are not intended to be printed. McLuhan described this phenomenon as though looking at the rear-view mirror; the old medium is always the content of the new medium, This is Marshall McLuhan: The Medium is the Message, McLuhan, 1967. While scholarly communication is transforming due to the effects of the Web, its content, shape, and way of communicating is still based on the characteristics of print media.

In Evolution of the Web, Berners-Lee, 1998, mentions the mandate to maintain the Web as an interoperable information space while evolving with the needs of society and technology. Interoperability meant that different systems without prior out-of-band knowledge should be able to communicate if they agree to operate based on open standards. Evolvability meant that languages and interfaces need to handle extensions, mixing, accommodate diverse scenarios for information exchange, as well as co-evolve.

The Web architecture initially included three independent components:

• the Universal Document Identifier (UDI) to refer to information on the Web – now known as Uniform Resource Identifier (URI) to identify things, and Uniform Resource Locators (URL) to locate them on the Web.
• the Hypertext Transfer Protocol (HTTP) as the request-response mechanism for systems – client–server model – to communicate and exchange hypermedia documents on the Web.
• the Hypertext Markup Language (HTML) as the lingua franca to describe and navigate hypertext and multimedia documents on the Web.

The W3C Technical Architecture Group (TAG) codified the design principles, constraints, and good practice notes of the Web architecture in W3C TAG Finding, Architecture of the World Wide Web, Volume One (AWWW), 2004. It documents the three architectural bases of the Web: Identification, Interaction, Data Formats, and states that the specifications emerging from each may evolve independently – according to the principle: Orthogonality – in order to increase the flexibility and robustness of the Web.

URI         = scheme ":" hier-part [ "?" query ] [ "#" fragment ]

The Web architecture initially used HTML as the language to represent and connect hypermedia documents with one another. HTML’s success is indisputable today, and that’s perhaps due to its simplicity to create and reuse; its flexibility, for partial understanding to proceed; and its extensibility to allow language mixing. While the HTML specification continues to evolve by reflecting on the needs of the Web, by design it is scoped to creating and connecting Web documents and applications.

Nevertheless, the potential language to describe the interrelationships between things as originally proposed in Information Management: A Proposal was not fully realised until the Resource Description Framework (RDF) came along. RDF is a constructed machine-interpretable language to describe and interlink arbitrary things at any level of abstraction on the Web, and reuses the existing architecture of the Web. In Sense and Reference on the Web, Halpin, 2009, states that in the context of RDF (and the Principle of Linking), URIs are primarily referential and may not lead to access unlike links in traditional hypertext systems.

Conceptually, RDF was in the fabric of the Web from the start. By naming things with URIs, we can construct structured sentences with RDF. By comparison, HTML at its core is a terse approach to representing and linking knowledge. RDF on the other hand is intended to be a unifying language for machine interfaces to data systems. I will further discuss language mixing with RDF in HTML and XML-family languages via RDFa.

One conceptual view of the Web is that anyone being (technically) allowed to say anything about anything (AAA). One consequence of AAA is that while systems (and reasoners within) can operate under both the open-world assumption (OWA) or the closed-world assumption (CWA), the Web ultimately uses the OWA framework because new knowledge can always make its way into the system, and lack of knowledge does not imply falsity. RDF applies a particular restriction of the AAA principle in that nonsensical, inconsistent, or conflicting information can be created or inferred. It is up to processing applications to determine their confidence on the information. Hence, an important distinction: assertions made using the RDF language are claims, as opposed to facts.

Statements in RDF are expressed in the form of triples – subject, predicate, object:

• the subject, which is an IRI or a blank node
• the predicate, which is an IRI
• the object, which is an IRI, a literal or a blank node

The Internationalized Resource Identifiers (IRI) is a generalisation of URI (US-ASCII) in that a wider character set (Unicode/ISO 10646) can be used to identify resources on the Web. A Blank node (anonymous resource) is a local identifier scoped to the context it is used in, eg. file, RDF store, and they are not intended to be persistent or portable. Literals are used for values like strings, numbers, and dates.

A set of triples in RDF (an RDF Graph) can be represented as a directed edge-labelled graph: $G=(V,L,E)$, where $V$ is the set of vertices (union of subject and object terms), $L$ is a set of edge labels (predicate terms), and $E$ contains triples from $V⨯L⨯V$. A subject may have several predicates, and predicates can also be mapped to subjects given that they have their own identity.

The structure of RDF statements is similar to:

• the subject–verb–object sentence structure in linguistic typology – used by nearly half of human languages
• the entity–attribute–value (EAV) model – widely used for advanced (meta)data modeling

We can represent the sentence "Otlet influenced Berners-Lee" in one of the RDF syntaxes as follows:

<http://dbpedia.org/resource/Paul_Otlet>
  <http://dbpedia.org/ontology/influenced> <http://dbpedia.org/resource/Tim_Berners-Lee> .

As ad hoc exploration is one of the goals of the Web, the W3C TAG Finding, The Self-Describing Web, 2009, reports how HTTP and other Web technologies can be used to create, deploy and access self-describing Web resource representations that can be correctly interpreted. It is expressed that the RDF language can be used to integrate with the Semantic Web such that the information encoded in a representation explicitly provides interoperable means to relate Web resources.

RDF facilitates a uniform follow your nose type of exploration by following the relations in statements in order to arrive at another unit of self-describing information. The uniformity in graph traversal enables applications to meaningfully interpret and integrate with disparate data without having any out-of-band knowledge, and retrieve more information about the terms as they need to from the source. As the content identified at http://dbpedia.org/resource/Paul_Otlet describes itself, in the same way, the relation http://dbpedia.org/ontology/influenced defines itself, and when we inspect a representation of http://dbpedia.org/resource/Tim_Berners-Lee, we will find that it describes itself.

We can describe Berners-Lee’s IRI (subject) by giving it a human-readable name (predicate):

<http://dbpedia.org/resource/Tim_Berners-Lee>
  <http://xmlns.com/foaf/0.1/name> "Tim Berners-Lee"@en .

In order to address different system and interface needs, different syntaxes for RDF emerged over time. RDF/XML is the grammar to define the XML syntax for RDF graphs; Turtle is a human-readable syntax resembling (subject–verb–object) sentence structure, RDFa expresses RDF statements in markup languages, JSON-LD serialises RDF in JSON (RFC 7159).

In order to facilitate querying and manipulating RDF graph content on the Web or in an RDF store, SPARQL Protocol and RDF Query Language (SPARQL) 1.1 provides a set of recommendations. The specifications address SPARQL 1.1 Query Language to match, filter, and construct graph patterns, as well as integration of disparate sources of information. SPARQL 1.1 Federated Querying is an extension to executing queries distributed over different SPARQL endpoints. SPARQL 1.1 Update to update, create, and remove RDF graphs in a store; a SPARQL 1.1 Protocol to convey queries and updates over HTTP; SPARQL 1.1 Service Description in order to discover SPARQL services, and a vocabulary for describing them. SPARQL 1.1 Entailment Regimes retrieve solutions that implicitly follow from the queried graph; and an alternative interface to SPARQL 1.1 Update uses SPARQL 1.1 Graph Store HTTP Protocol for operations on RDF graphs.

Using the example RDF data from earlier about Otlet’s relation to Berners-Lee and his name in an RDF store, the following query returns the data in a tabular format for the names of objects that match the graph pattern where the subjects influenced.

SELECT ?object ?name
WHERE {
  <http://dbpedia.org/resource/Paul_Otlet>
    <http://dbpedia.org/ontology/influenced> ?object .
  ?object
    <http://xmlns.com/foaf/0.1/name> ?name .
}

The SPARQL suite is a powerful set of tools for inspecting and manipulating information based on the graph nature of the data.

The Linked Data design principles build on the AWWW in order to identify, describe, and discover information, enabling a Semantic Web where humans and machines can explore the Web of data:

1. Use URIs as names for things
2. Use HTTP URIs so that people can look up those names.
3. When someone looks up a URI, provide useful information, using the standards (RDF*, SPARQL)
4. Include links to other URIs. so that they can discover more things.

In practice, what URIs identify, based on the resources fetched upon dereferencing, was ambiguous until the release of the famous W3C TAG Finding, httpRange-14 issue, 2007, as per httpRange-14: What is the range of the HTTP dereference function?. The core issue had to do with machine agents having a deterministic mechanism to distinguish between accessible digital objects on the Web (information resources) and references to a class of physical entities and abstract concepts (non-information resources), ie. things that are not technically "retrievable" over the wire on the Web, but may only be referred to. In Sense and Reference on the Web, Halpin posits that the definition of information object and information realization can be thought of as the classic division in philosophy of mind between an object given on a level of abstraction and some concrete thing that realizes that abstraction, where a single abstraction may have multiple realizations.

Having established an overview of the technical environment I will next discuss the Web Science framework, which situates the technical aspects alongside social processes.

Web science – what makes the Web what it is, how it evolves and will evolve, what are the scenarios that could kill it or change it in ways that would be detrimental to its use.

The openness of the Web and the ease of involvement for both humans and machines (through common protocols) helped its great expansion without a particular central point of failure or coordination. Scholarly communication could be exemplary of different kinds of agents interacting on the Web, whether they are sentient or something else. Web Science – Creating a Science of the Web – is an interdisciplinary field of study concerned with understanding and developing sociotechnical systems like the Web alongside human society. In A Framework for Web Science, Berners-Lee, 2006, the authors state:

the Web perhaps more than any other recent human construct carries with it any number of issues including privacy and protection, access and diversity, control and freedom. Structures that we design, engineer and research, and findings that emerge through analysis, will often have strong societal implications.

The challenges of Web Science comprise both social and technical aspects, from user interfaces to data, to information policy, resilience, access from diverse devices, collective creativity, and decentralisation. All of these areas are pertinent to the future of scholarly communication, and Web Science as a field provides a framework for ensuring we consider the various social and technical issues as part of an interconnected ecosystem.

In the same vein, furthering the study of Web Science itself can be done through advancements in how we communicate scholarly findings; through better enabling universal access and connections within our collective knowledge.

The Web not only expedited the flow of human knowledge with a friendly interface, it enabled social interactions to take place which would not have been possible otherwise due to physical or social constraints. The global adoption of the Web brought new forms of human association and social action. In essence, the Web creates one form of a global village, described by McLuhan, whereby there is greater requirement of individuals to participate than with earlier forms of electric media eg. telegraph, electric light, telephone, radio, television. The way society reacts to the Web as a medium is perhaps more important than the content that is on it.

In Credibility of the Web: Why We Need Dialectical Reading, Bruce, 2000 explores Kaufmann’s modes of reading from the 1977 essay Art of Reading (The Future of the Humanities) in context of the characteristics of the Web:

• exegetical: author is authority, the reader is passive;
• dogmatic: scepticism on part of the reader;
• agnostic: continuous evaluation while acknowledging good and bad qualities;
• dialectical: possessing coding, semantic, and pragmatic competence

Bruce draws attention to the dialectical in that an observer enters into a deep experience of reading, engaging with the text with a critical eye, drawing advanced information and interpretations beyond what is merely presented, and actively seeking to understand the material in addition to political, social, and historical dimensions. Bruce states that under the dialectical view, the Web’s multimodal features allows new values and ways of making meaning as an holistic involvement for the consumer. This view is a useful exemplification of McLuhan’s notions on the effects of media on society.

While there are many definitions for “open” and “free”, depending on the context assigned by different communities eg. The Many Meanings of Open, Berners-Lee, 2013, here I acknowledge some that are commonly used.

The definition given by The Open Definition which is derived from the Open Source Definition, is as follows:

Knowledge is open if anyone is free to access, use, modify, and share it — subject, at most, to measures that preserve provenance and openness.

The definition of free (or libre) in context of software that is widely acknowledged is:

“Free software” means software that respects users’ freedom and community. Roughly, it means that the users have the freedom to run, copy, distribute, study, change and improve the software. Thus, “free software” is a matter of liberty, not price. To understand the concept, you should think of “free” as in “free speech,” not as in “free beer”. We sometimes call it “libre software,” borrowing the French or Spanish word for “free” as in freedom, to show we do not mean the software is gratis.

In order for an item or piece of knowledge to be transferred in a way to foster rich culture, the Free Cultural Works defines freedom to mean the freedom to use; to study; to make and redistribute copies; and to make changes and improvements. The definition can be used as a tool to determine whether a work or license should be considered "free."

The definition of Open Content explains how users of works to have free and perpetual permission to engage in activities with the right to retain, reuse, revise, remix and redistribute content.

Creative Commons’s (CC) CC licenses is one of public copyright licenses that can be used on works which grant additional permissions for the purpose of free distribution of an otherwise copyrighted work. With the exception of CC0 1.0 (public domain) license, all CC licenses require users of the works to attribute the creators. CC0, CC BY (Attribution) and CC BY-SA (Attribution-ShareAlike) are compatible with the definition of Free Cultural works. CC licenses are open, flexible, interoperable, and has global uptake. The policy of CC license is also compatible with public laws in various jurisdictions.

In 1991, recognising the need for scholars to expedite the exchange of scholarly information, Paul Ginsparg created the LANL preprint archive for e-prints. Creating a global knowledge network, Ginsparg, 2001, states that self-archiving provided a moderated central repository where scientists can deposit their articles to make them accessible to anyone, which was a cost-effective solution (estimates were less than 5 USD per submission for administration) for scholars to disseminate research results quickly. It initially started out as a central mailbox with an interface; it adopted FTP in 1991, Gopher in 1992, and finally the World Wide Web in 1994. The articles in the repository are not expected to be peer reviewed, but rather a place where versioned preprints – considered to precede publications that are peer reviewed – can be accessed for free. In 2001, the LANL preprint archive eventually became what is currently known as arXiv. The arXiv license information permits different rights and licenses to be assigned to the scholarly records, ie. anything from public domain, Creative Commons licenses, or with copyright to the publisher or author, provided that arXiv is given sufficient rights to distribute. The repository accepts submissions in TeX, PDF, PostScript, and HTML. This sort of discipline-centric and institution-run preprint exchange mechanism had the technical ingredients to serve as a substitute for the traditional scholarly journal. From 1991 to 2017, arXiv had a linear increase in submission rate (passing the 1 million mark in 2015). arXiv’s early success lead to the emergence of the Open Archives Initiative in 1999, and the Open Access movement in 2002.

In 1994, Harnad made a proposal to appeal to the esoteric:

It is applicable only to ESOTERIC (non-trade, no-market) scientific and scholarly publication … that body of work for which the author does not and never has expected to SELL the words. The scholarly author wants only to PUBLISH them, that is, to reach the eyes and minds of peers.

After some years of prototyping and researcher-centric initiatives, the ideas eventually formulated into the Budapest Open Access Initiative (BOAI) in 2002. BOAI is based on the principle that publicly funded research output, including peer reviewed journal articles as well as unreviewed preprints, should be made immediately available, and indefinitely, without access fees and free from most restrictions on copying and reuse, to increase visibility and uptake:

By "open access" to this literature, we mean its free availability on the public internet, permitting any users to read, download, copy, distribute, print, search, or link to the full texts of these articles, crawl them for indexing, pass them as data to software, or use them for any other lawful purpose, without financial, legal, or technical barriers other than those inseparable from gaining access to the internet itself … the only role for copyright in this domain, should be to give authors control over the integrity of their work and the right to be properly acknowledged and cited.

In order to open up access to scholarly journal literature, BOAI recommends the adoption of two complementary strategies for open access (OA):

• Self-Archiving
• Open-access Journals

The term self-archiving in context of OA differs from that of the early practices around preprint self-archiving in that there were no constraints set for the status of the scholarly literature, ie. peer reviewed or not. For OA-centric scientific literature, quality-control and certification are integral.

The foundations of the OA approach were conceived at a time when the major revenue models to cover publishing costs for scholarly literature were through subsidies from institutions, journal subscriptions, or pay-per-view. In this payment model, consumers of the literature were charged either individually or through institutions. In contrast, the article processing charge (APC) model requires publication costs to be covered upfront, thereby waiving the costs for the consumers of the literature. Today, APC is the most widely used method to finance OA publishing, with varying levels of fees. The OpenAPC dataset includes publication fee spending of 238 institutions reveals an average of 1946 EUR per article based on fully and hybrid OA journals (average of 1558 EUR and 2504 EUR respectively). In this dataset, the institutional spending on article publishing in 2018 was at minimum 13 EUR, maximum 12000 EUR, and mean 1834 EUR.

The effect of open access and downloads ('hits') on citation impact: a bibliography of studies, OpCit, 2013, is intended to describe progress in reporting these studies; it also lists the Web tools available to measure impact. It is a focused bibliography, on the relationship between impact and access. Harnad argues that the essential message of OA is research that is not freely available online loses research impact – any reasons not to make research Open Access are merely excuses (see the raincoat science metaphor).

In The deliverance of open access books, Snijder, 2019, looks at the consequences of open access for books, and concludes that books that are freely accessible online attract more readers and are cited ten percent more than those that do not use open access platforms.

In Modularity: the next form of scientific information presentation?, Kircz, 1997, states that the present-day linear (essay-type) scientific article is the result of a development over the centuries, and that different types of information, at present intermingled in the linear article, can be separated and stored in well-defined, cognitive, textual modules. A natural consequence to having modular electronic articles, with their own unique characteristics, is that units are self-contained and interconnected to other units.

In The deconstructed journal – a new model for academic publishing, Smith, 1999, provides insights on networked-based scholarly publishing models:

1. means/end confusion, where journals continue to mimic or replicate existing mechanisms
2. modifications to the system should still match the needs of the traditional scholarly journal
3. cooperating agents can function without going through a central agency, eg. a publisher.

The role of the journal includes both main (editorial, quality control, recognition of work, marketing, and disseminating) and hidden (subject and community defining, and archiving) functions. The proposed deconstructed journal model is preferable to the paper influenced electronic publishing model in that both main and hidden roles can be accomplished in a distributed manner. Here, Smith emphasises allowing greater academic freedom and shifting of power and control from the monolithic third-parties to the knowledge producers, which could be realised in either an institution- or researcher-centric approach. In this process, journals with the refereeing process would link to the articles with full content hosted on the authors’ Web sites. To get there, however, Smith acknowledges that for such a new paradigm to be adopted, professional and funding bodies must transition themselves to accepting the decoupling and self-controlled literature publishing model as something equivalent (in terms of academic value) to the traditional model.

After two decades, the anecdotal evidence suggests that academia acknowledges the sort of equivalence that Smith describes at varying levels. For example, while literature quality-control can be generally anywhere on the Web, on the other hand, self-published literature remains to be treated as grey literature, that is, any material that is outside of the traditional scholarly distribution channels, especially when it comes to academic reward systems.

In Decoupling the scholarly journal, Priem, 2012, presents an analysis of the traditional functions of the scholarly journal – registration (recording, time-stamping), certification (quality-control, assessment), awareness (dissemination), and archiving (preservation) will be discussed in detail under Forces and Functions. The article echoes earlier work in that the scholarly journals bundle multiple functions, and that this tight coupling has hindered innovation, is inefficient, and makes it difficult to improve individual functions. The article also states that the current system, ie. the bundling of the closed publishing and certification models, is unlikely to change if the communication functions are not decoupled, regardless of all the other out-of-band activities – activism or innovation – to improve the publishing models or certification mechanisms. The article diverges from the traditional view of separating registration and archiving functions since the latter necessitates both persistent storage and identification. Hence, an archived unit of communication (Rethinking Scholarly Communication) has already fulfilled registration. The authors echo many of the ideas of Smith with respect to decoupling and decentralisation, however they only go as far as describing them within the context of scholarly functions operating alongside institutional or third-party online services. While the article does not exclude the possibility of units of information being self-hosted and individually-controlled by the researchers, it is not explored further. The article concludes that the nature of the communication revolution will be through the structure and organisation of the units in the system.

The First meeting of the Open Archives initiative in 1999 was held to agree on recommendations to address interoperability challenges, and outlined in The UPS Prototype, 2000 – such as search and linking – for data across publicly available distributed self-archiving solutions. To promote the global acceptance of self-archiving, with interoperable systems being core to its successful adoption, the Open Archives Initiative (OAI) Organization was established to take on the development of technical recommendations – for metadata standards, discovery, and retrieval. This led to the evolution of recommendations such as OAI-PMH Protocol for Metadata Harvesting, and its successor ResourceSync as a resource synchronization framework for servers; OAI-ORE for Object Reuse and Exchange; and Memento which specifies a time dimension to the Web architecture. Reminiscing About 15 Years of Interoperability Efforts, Van de Sompel, 2015, summarises the lessons learned on interoperability efforts by concluding that a shift from a repository-centric to a Web-centric perspective was essential in order have a viable and sustainable scholarly system.

While there is a plethora of overlapping guides, best practices, principles, manifestos, and declarations in the field of research, scholarly communication, data and publishing, here I highlight a few that are Web-centric.

Today, Open Science is a broadly understood notion that is intended to foster an accessible approach to conducting science and scholarship. While there is no single agreed definition of open science, the general consensus is about improving the practice of scientific communication in a transparent, traceable, verifiable, and reusable manner. One definition of open science is as follows:

Open Science is the practice of science in such a way that others can collaborate and contribute, where research data, lab notes and other research processes are freely available, under terms that enable reuse, redistribution and reproduction of the research and its underlying data and methods.

With the advent of the scholarly journal, the practice of open science have been continuously evolving. As the Web speeds up information exchange, there has been further interest in the systemisation and application of open science in research communities. Later on I will further discuss Social Scholarly Web.

Data on the Web Best Practices

The W3C Recommendation for Data on the Web Best Practices (DWBP) provides a wide range of best practices, in order to benefit comprehension, processability, discoverability, reuse, trust, linkability, access, and interoperability. The DWBP can be used by data publishers in order to help them and data consumers to overcome the different challenges faced when publishing and consuming data on the Web. DWBP applies to all data publishers but is pertinent for researchers.

FAIR Guiding Principles

The FAIR Guiding Principles, Wilkinson 2016, is a set of principles to facilitate knowledge discovery and reuse for machines and humans. The principles, findable, accessible, interoperable, reusable (FAIR) are intended to express policy goals, as opposed to a technical prescription to building a data infrastructure and exchange. The elements of the principles include:

Findable

For human and machine discovery of (meta)data and services by use of globally unique and persistent identifiers, and machine-readable way-finding.

Accessible

(Meta)data to be retrievable using open and free standard communication protocols via applicable authentication and authorization mechanisms.

Interoperable

(Meta)data using structured knowledge representations to be interpretable by autonomous agents.

Reusable

(Meta)data to optimise the reuse of data with domain-relevant, license, and provenance information.

Cloudy, increasingly FAIR; revisiting the FAIR Data guiding principles for the European Open Science Cloud, Barend, 2017, posits that data are often open but not necessarily FAIR; data could be closed yet perfectly FAIR. FAIR does not directly proscribe data quality, trustworthiness, ethics or responsibility. Hence, conforming with the FAIR principles does not necessarily adhere to the scientific method.

Cost of not having FAIR research data, PwC EU Services, European Commission, 2018, provides a cost-benefit analysis for FAIR research data based on the following indicators: time spent, cost of storage, license costs, research retraction, double funding, cross-fertilization, potential economic growth (as % of GDP). The estimates reveal that the annual cost of not having FAIR research data costs the European economy at least €10.2bn every year and that the true cost cost to be much higher. To put €10.2 billion in perspective, the cost of not having FAIR is ~ 400%, of what the European Research Council and European research infrastructures receive combined. The study also highlights the consequences in the absence of FAIR data to include impact on research quality, economic turnover, or machine readability of research data.

Data Processing, Privacy and Protection

In addition to institutional ethical principles and federal regulations, research can be subject to additional regulations. For instance, the General Data Protection Regulation (GDPR) imposes obligations on organisations that process personal data of individuals in the EU and the EEA. Hence, GDPR is relevant for experiments and studies that include human subjects, personally identifiable information (PII), as well as acquired or inferred personal characteristics. In relation to scholarly communication, ie. research as a basis for data processing, GDPR is permissive and indicates exemptions on the processing of personal data for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes should be subject to appropriate safeguards for the rights and freedoms of the data subject, as well as for scientific research purposes including technological development and demonstration, fundamental research, applied research and privately funded research.

Digital Agenda

In order to further open up the market for services based on public-sector information, the European Commission’s Revision of the PSI Directive, 2012 — a digital agenda for Europe — as per Office Journal C 240/2014, recommends high-demand datasets from libraries and archives be published in machine-readable and open formats (CSV, JSON, XML, RDF, etc.) to enhance accessibility, and described in rich metadata formats and classified according to standard vocabularies, and to facilitate re-use, public sector bodies should … make documents available … at the best level of precision and granularity, in a format that ensures interoperability.

As the importance of the new approaches are acknowledged by scholarly communities, there is an increasing social, as well as technical demands for their adoption.

Because paper enforces single sequence and there is no room for digression, it imposes a particular kind of order in the very nature of the structure. When I saw the computer, I said, ‘at last we can escape from the prison of paper’, and that was what my whole hypertext idea was about in 1960 and since. Contrarily, what did the other people do, they imitated paper, which to me seems totally insane.

The Web goes beyond the confines of the physical paper. The user experiences between the Web and print-centric media differ greatly. This section briefly compares Portable Document Format (PDF) and HTML.

Donald E. Knuth, the inventor of TeX, considers TeX to be a program written using the literate programming approach. The TeX source code can be imperatively programmed, then compiled by a processor to arrive at a view. LaTeX, which is widely used for scholarly articles, is a collection of TeX macros. In some scientific communities, PDFs are usually generated from LaTeX.

PDF (ISO 32000-2:2017) is intended for displaying and storing, and generally self-contained. However, it is not intended to be editable as the formatting instructions are no longer available. PDF is inherently layout oriented, and it is an optimal format for printing given its high-precision for typography. Third-party scholarly publishers tend to require a fixed layout with typographical guidelines for research documents, hence submissions are often required to be made as PDFs.

In order to facilitate metadata interchange, an Extensible Metadata Platform (XMP) package in the form of XML (most commonly serialized as RDF/XML) may be embedded in PDF as per XMP Specification Part 1: Data Model, Serialization, and Core Properties (ISO 16684-1:2012). Otherwise, semantically useful information is not preserved when PDFs are generated, which makes it difficult to go back to source content format.

HTML is both a content format for encoding information, and a document format for storing information. HTML can reflow to fit different displays. HTML has ubiquitous support across devices.

If HTML is used as the original format to record the information, a PDF copy can be generated on demand, using the specifications of the user. With an accompanying CSS for the HTML, desired layout guidelines can be achieved. Converting from PDF to HTML+CSS is possible, but the precision rests on rendering the view, as opposed to creating a structurally and semantically meaningful document.

Formatting research articles as PDF comes at the cost of losing access to granular information. Thus, PDF’s primary focus – the presentation layer – leads to legacy documents that are not reusable in the long run. A great amount of reverse-engineering is required to extract the underlying information (like content, tables, figures). The source format (like TeX, JATS, XML, HTML, or other) are often needed instead for content exchange and modifications.

TeX and HTML (as stacks) are compared in Table Comparison of TeX and HTML.

Third-party research publishing services commonly impose constraints on file formats derived from print requirements. As a consequence, it regulates researchers’ choice of tools to prepare and share their work with their peers and the society. In essence, it impacts how researchers and scholars approach communicating their work based on the confines of the medium. While the choices for digital representation for research information may seem arbitrary, the effects of a print-centric scholarly publishing is analogous to McLuhan’s we look at the present through a rear-view mirror.

I chose HTML not to be a programming language because I wanted different programs to do different things with it: present it differently, extract tables of contents, index it, and so on.

One of the shortcomings of the research publication system is that scientific literature in electronic formats resemble, mimic, and have similar affordances to printed papers. While the paper-centric approach has been adequate, although not necessary optimal, for human consumption and reuse, it is insufficient for machine-assisted analysis or reuse. The print-centric affordances lack structure that could enable better aggregation, discovery, processing, and integration – which machines excel at – and in turn help with human-led tasks. Representing and storing semantically structured knowledge on the Web enables automated agents to process information more intelligently.

The most wide use of the term semantic publishing in scientific literature is defined in Semantic publishing: the coming revolution in scientific journal publishing, Shotton, 2009. Here the term semantic publishing is constrained to anything that enhances the meaning of a published journal article, facilitates its automated discovery, enables its linking to semantically related articles, provides access to data within the article in actionable form, or facilitates integration of data between papers. In Genuine Semantic Publishing, Kuhn, 2017, posits that with respect to the Semantic Web vision, this definition is too restrictive in a sense that it is concerned mainly about narrated information, and too inclusive in that a shallow semantic publication would still qualify. The article argues that the definition for semantic publishing should be broader in that works without a narrative are accounted for, and narrower in the sense that semantic representations should be incorporated at the time of creating and publishing entities. The genuine semantic publishing concept prescribes the notion of publishing 1) machine interpretable formal representations, 2) having essential coverage of the claims, 3) authenticated by the contributors, 4) be a primary component of the work, and 5) be part of fine-grained and light-weight containers.

Analogous to the FAIR principles, data needs to "signal" potential opportunities for reuse as per Research Objects in Why Linked Data is Not Enough for Scientists, Bechhofer, 2013.

A core component of the Web is hyperlinks between documents. Most, if not all, academic articles reference others in some way. For articles published using Web technologies for the content or even just the metadata, creating hyperlinks is an obvious way to enhance these references. Such links can create citation chains between articles at a high level, and graphs of claims and arguments at a more granular level.

The case is made for using a Web-friendly and machine-tractable model of scientific publications in Micropublications: a semantic model for claims, evidence, arguments and annotations in biomedical communications, Clark, 2014. The Micropublications model, formalised as an OWL ontology, can reduce the human-labour cost of checking the support of a claim, as it is capable of representing scientific controversy and evidentiary requirements in primary literature. Further, the Micropublications model can be applied to generate argument graphs representing the claim that is argued, and its support by other statements, references, and scientific evidence (data). Machine-navigable linkage at the level of assertions in scientific articles and data improves robustness of scientific citations and observations. Argument graphs: Literature-Data Integration for Robust and Reproducible Science, Clark, 2015, posts that creating data structures based on the relationships between arguments can be complementary to using entity recognition for mapping textual terms in articles to curated scientific databases.

Citations perform a variety of functions, and an author’s intention when referencing a particular source is not always evident from the prose, and certainly not from a machine-readability perspective. For example, does the author mean to point to a whole document (this is usually the level at which traditional citations operate) or are they only referring to a particular statement or section within it? Are they citing something because it supports their perspective, or because they disagree with it? This is particularly problematic in a system which simply counts the instances of a citation to determine impact or relevance. Measuring academic influence: Not all citations are equal, Turney, 2015, and Why (almost) Everything We Know About Citations is Wrong: Evidence from Authors, Teplitskiy, 2018, provide evidence indicating that authors with a means to express their citations and intentions in a structured way can improve the preservation of the connections between research assertions or units of information in general.

Fortunately there has already been work towards creating vocabularies around the relationships between articles and their citations. In An annotation scheme for citation function, Teufel, 2006, a classification system is proposed for relationship information between articles and their citations in order to improve automatic recognition to build better citation indexers, as well as to have more accurate interpretations of scientific arguments.

The Citation Typing Ontology (CiTO) model is equipped with common classifications eg. "cites for information", "disputes", "cites as evidence", "cites as potential solution", that can be used to create explicit relations between entities, as posited in FaBiO and CiTO: ontologies for describing bibliographic resources and citations, Peroni, 2012. It also has an extension mechanism to support customised properties in order to allow authors to express their intentions closer to their mental model.

In contrast to extracting rhetorical knowledge structured claimed by authors of scientific articles, Corpora for the conceptualisation and zoning of scientific papers, Liakata, 2010, describes how the structure of human-readable investigation in research articles can be retrieved based on a generic high-level annotation scheme for scientific concepts. It is shown that such complementary approach can assist in the generation of automatic summaries and intelligent querying.

Genuine Semantic Publishing, Kuhn, 2017, also contends that while narrative text remains an important part of research discourse, publishing the data using formal semantics without the need of a narrative is beneficial. Meanwhile, tooling for creating unique identifiers at granular levels, for clauses or statements within articles as well as pieces of supporting datasets, hinders researchers’ in using structured mechanisms for describing their citations. Authors may be unable to obtain or generate a precise identifier, or lack an authoring or referencing environment which would allow them to assign identifiers appropriately. When an author’s intentions are preserved only in prose, entity-recognition techniques are typically required to understand the original intentions of citations. This is generally a complicated task and error prone given that an algorithm attempts to reverse-engineer the purpose of the citation.

Identifying units on the (scholarly) Web is currently done with different kinds of social agreements and expectations. In Identifiers for the 21st century: How to design, provision, and reuse persistent identifiers to maximize utility and impact of life science data, McMurry, 2018, posits that actors should generally create their own identifiers for new knowledge, and to reuse existing identifiers when referring existing knowledge. Depending on the one or more roles an actor has in the creation, editing, and republishing of content, the decision to create a new or reusing identifier is left as a judgement call.

Here I discuss a few systems that pertain to registration of identifiers:

Digital Object Identifier (DOI) is an international standard for document identification, and the services that are offered through doi.org is managed by the International DOI Foundation (at the time of this writing). The foundation permits DOI Registration Agencies to manage DOI records. In order to have a resource assigned with a DOI, it has to be done by one of the DOI registration agencies which requires organisational membership, excluding individuals. For example, a typical journal publisher as a member of a DOI registration agency (eg. CrossRef) submits mappings for objects of interest. Given the statement in AWWW, URI persistence is a matter of policy and commitment on the part of the URI owner, DOI introduces an additional layer of social contract to govern the discovery of research artifacts. A DOI is aimed at being a persistent identifier for an object it is assigned to, and has the mechanism to refer to the URL of an object where it can be found. That is, a DOI may refer to a landing page about the actual object and not necessarily the object itself – this happens to be a common practice for literature that is not necessarily open and free. For literature that has restricted access, the resolved DOI is typically a webpage where an actor would need to go through the system’s paywall or access toll in order to retrieve the content. From a machine’s perspective, what the DOI identifies at the resolved location is not clear eg. company logo, tracking software, search form or the intended research object? For open literature, the resolved DOI would generally hyperlink to the actual scientific object in human-interpretable only fashion, or be part of a webpage accompanied by material unrelated to the scientific object itself, ie. information that the scientists neither assembled as part of their research nor intended for the consumer. If the actual URL of the object changes, the DOI’s lookup table for the URL has to be updated. Otherwise, they'll point to dead links or unrelated information.

Open Researcher and Contributor ID (ORCID) is a community-governed persistent identifier system for researchers and organisations. An ORCID identifier is specified as a URI. While anyone can create and update their ORCID profile, the identifier is ultimately owned and managed by the ORCID organisation and made available through orcid.org. Profiles are constrained to express to the UI’s template and the kind of information ORCID is willing to store. For most intents and purpose, the free service offered by ORCID is sufficient for researchers, and provides a mechanism where researchers can aggregate information about their education, employment, funding, and works like academic publications and data. It also allows researchers to extend their profile descriptions, as well as linking to their alternative online profiles.

Persistent Uniform Resource Locator (PURL) is a way to provide an HTTP URL a permanent redirect to another HTTP URL. PURLs can be registered and managed by anyone. The PURL service is currently administered by the Internet Archive.

Similar to PURLs, is the W3C Permanent Identifier Community Group’s resulting w3id.org, where a group of entities that pledge to ensure the operation of the community-drive service. w3id.org provides a permanent URL re-direction service, and it can be done by making a request to register and update a URI path. There are shared second-level paths eg. for people, that can be registered the same way. The service is flexible in that different server rules for redirections can be set.

Persistent Identifiers for Scholarly Assets and the Web: The Need for an Unambiguous Mapping Van de Sompel, 2014, states that a Persistent Identifier (PID) is a string that complies with a well-defined syntax, minted in a namespace controlled by a naming authority. PURL, w3id, ORCID, and DOI are PIDs. For digital objects accessible over the Internet, PIDs are intended to be used as a long-lasting reference to a unit of information, and be actionable in the sense that the environment can be directed to the identified source. For example, a DOI-based PID can be 10.2218/ijdc.v9i1.320, and the organisation – in this case the International DOI Foundation – would be responsible to maintain an unambiguous mapping to an accessible resource. Typically once mapped, the process starts from a DOI URI https://doi.org/10.2218/ijdc.v9i1.320 which (currently) resolves to the location http://www.ijdc.net/article/view/9.1.331 where its contents can be retrieved. URN, International Standard Book Number (ISBN), Archival Resource Keys (ARK) are other examples of PID.

There are different forms of social contracts or promises made for long-term societal benefit in terms of archival and preservation of knowledge. They can be categorised as follows:

• any Web-wide publicly usable archival services, eg. Internet Archive, archive.is, WebCite, Perma.cc, Webrecorder;
• dedicated digital preservation organisations, eg. Portico;
• libraries, eg. LOCKSS;
• global archives preserving content on behalf of all libraries, eg. CLOCKSS;
• subscription based service for all kinds of libraries, federal institutions, state archives, NGOs, eg. Archive-It;
• state or federal archives, eg. Swiss Federal Archives, Library and Archives Canada;
• institutional-run digital archives, eg. TIB, Zenodo

In 1996, the Internet Archive begun to archive the publicly accessible resources on the Internet, and provides free public access to its digital library. The Wayback Machine is a service that allows users to search and access archived Web documents in particular. The Wayback Machine can be used to navigate through different versions of URLs based on the date-time in which they were accessed and stored. For example, all of the captures of csarven.ca as well as the other articles of the site are available, along with a summary.

Contemporary social media and networking sites has been facilitating academics to exchange information. More specifically, academics use social media for a variety of reasons, to name a few: to network and collaborate; disseminate and find research results, crowd-source and crowd-fund research challenges, engage with public and foster trust on research results by providing authoritative feedback.

There are online communities geared towards scientists, academics, librarians, and scholars with topical focus on various aspects of scholarly communication. In Academics and their online networks: Exploring the role of academic social networking sites, Jordan, 2015, discusses the proliferation of social networking sites predominantly operated by businesses independently of education institutions. Academics from different disciplines communicate to publish and review articles in different states (pre-prints to post-prints), pose and answer questions, construct identities and build profiles, exchange messages, as well as explore trends.

Online social media tools are affecting or influencing scholarly communication from a variety of fronts. For instance, with the advent of the Web and the notion academics having the right and means to exchange scholarly literature with their peers for free, there has been plethora of initiatives and developments over the course of three decades in academia, libraries, archives, as well as standards for creating, sharing and preserving (Open Access Directory). Moreover, the call to take advantage of features offered by the new media, has opened new ways of publishing and disseminating scientific and scholarly knowledge globally, as well as institutional infrastructures being adapted to meet communication needs. For example, universities providing an online space for staff members and students to publish their profiles, as well as to share their research interests, academic output, and other educational material. Institutional repositories have been built to retain copies of scholarly work, as well as mechanisms to exchange data with other repositories.

In recent years, there has also been growing number of community-led initiatives for individuals to help researchers and scientists by performing citizen science independently, from sharing photos of coral reefs, analysing radio signals, to solving protein structures.

In parallel to "internal" technical and social advancements, infrastructure towards open scholarly communication has been offered by commercial entities in various forms with different business models, from authoring, reviewing, messaging, organising, indexing, searching, to visualising scholarly information.

In essence, third-party controlled online social media exist simply because academics have needs to be fulfilled. PASTEUR4OA Briefing Paper: Infrastructures for Open Scholarly Communication, Moore, 2016, states that online platforms have become increasingly centralised locations for a range of user interactions, and they have a commonality in that, their primary currency is user’s data which is monetised in various ways; the platforms cater to different stakeholders while being free; the platforms are often funded by firms that offer venture-capital and are almost exclusively driven by profit-maximising; and in order to maximise the number of users and data generated through their interactions, the platforms are centralised and act as closed proprietary ecosystems (commonly referred to as walled gardens) where data access, portability, ownership and content structures are constrained in ways which makes it difficult for users’ to simply move to another system.

peer review

A method of quality control of scholarly articles, whereby each article submitted for publication in a journal is sent to one or more scholars working the same research field as the author and the scholar(s) assess(es) whether the article is of a high enough standard and appropriate for publication in that particular journal.

In Free at Last: The Future of Peer-Reviewed Journals, Harnad, 1999, states that peer review is a quality-control and certification (QC/C) filter necessitated by the vast scale of learned research today. In Implementing Peer Review on the Net: Scientific Quality Control in Scholarly Electronic Journals, Harnad, 1996, posits that electronic networks offer the possibility of efficient and equitable peer review on published and ongoing work, which can replicate the traditional paper form. As the medium revolutionised interactive publications, it was possible for peer review to be supplemented with online commentary as another form of quality control for scholarly literature at any state.

The scientific community has been aware of the crude and fallibility of traditional peer review, for example, in cases such as, crediting bad research, good research getting dismissed for political reasons, predatory publishing (quality of the research content being irrelevant), or even treating commentary or opinion pieces as equivalent to scientific evaluation. Nevertheless, the function of peer review in the scholarly system is a form of self-regulation and certification. Without having any maintenance for quality, distinguishing valid, reliable, useful and ethical work from those not, would be difficult to achieve systematically. This is particularly concerning where the general public can become aware of a piece of information, yet the validity of the research findings may be inaccessible or absent, and thus have major consequences to our society and the planet – as with cases where widely agreed scientific evidence and observable effects of global climate change can be easily "questioned" by propaganda spread through mass-media by elite forces with a personal or private agenda; the amplification of the flat Earth conspiracy, and so on.

What is open peer review? A systematic review, Ross-Hellauer, 2017, first categorises the problems in traditional peer review, second studies the diversity of definitions of open peer review (OPR) in literature, and sets out to resolve the ambiguity, and classifies the common traits. The articles argues that the OPR traits: open identities, open reports, open participation, open interaction, open pre-review manuscripts, open final-version commenting, and open platforms (“decoupled review”) can help to address common problems in the traditional peer review model categorised as: unreliability and inconsistency, delay and expense, lack of accountability and risks of subversion, social and publication biases, lack of incentives, wastefulness.

The peer review model towards quality control has received extensive criticism. These are studied in A multi-disciplinary perspective on emergent and future innovations in peer review, Tennant, 2017. The article examines different approaches to peer review models and systems. In context of decoupled post-publication (annotation services), the article states that while there are online communities and services for open peer review, the systems are largely non-interoperable in that most evaluations are difficult to discover, lost, or rarely available in an appropriate context or platform for re-use.

Reproducible research findings refers to the notion that when the exact methods are used to replicate or repeat to reach the same findings. However, for some research, reproduction is not always possible, for example when data is sensitive, or influenced by time or space. In this case, replication can be carried out to answer the same research questions using similar or equivalent approaches. The reproduction of the original work can be either confirmatory, ie. able to reproduce, or; contradictory, unable to reproduce or inconclusive. Thus, in order to reconstruct how the data was generated and to make sense of it, data needs to be accompanied by enough detail about its provisional interpretations.

The preregistration revolution, Nosek, 2018, describes the uptake of the preregistration process as a way to define the research questions and analysis plan before observing the research outcomes. Registered reports was initially proposed as an open letter to the scientific community as a way to Changing the culture of scientific publishing from within , Chambers, 2012. Researchers would register a research plan which gets reviewed by their peers for significance, rationale, soundness and feasibility of the methodology, and replicability, it would be approved to carry out the research. The approach provides a conditional acceptance for the final manuscript.

Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.

The current scholarly system implements metrics for impact of various scholarly artifacts. Citations impacts quantifies the importance of academic articles, journals, and authors. Article citations is used to indicate how often it is referenced. Journal Impact Factor (JIF) is used to indicate the relative importance of an academic journal within its field. h-index is one way to measure bibliometric impact of individual authors in their field. The validity of these metrics has received numerous criticisms.

Prestigious Science Journals Struggle to Reach Even Average Reliability, Brembs, 2018, examines accumulating data which suggests methodological quality and, consequently, reliability of published research works in several fields may be decreasing with increasing journal rank. As society becomes more literate and reliability of research becomes more important, the article posits that the scientific community should strive to abolish the pressure that JIF exerts towards measuring the productivity of scholars, in order to preserve society’s trust in the scientific endeavour. Deep impact: unintended consequences of journal rank, Brembs, 2013, posits that fragmentation and the resulting lack of access and interoperability are among the main underlying reasons why journal rank has not yet been replaced by more scientific evaluation options, despite widespread access to article-level metrics today.

Altmetrics is the creation and study of new metrics based on the social web for analyzing, and informing scholarship. They are are complementary to traditional citation impact metrics by aggregating diverse online research output eg. citations, downloads, mentions, from public APIs (The Altmetrics Collection, Priem, 2012). It could be applied in researcher evaluations, as well as can pressure individuals to use commercially controlled social media.

The scholarly communication system has been the subject of controversies internationally ever since the shift towards its systematisation and the effect of the forces in society, research market and knowledge industry. Controlling access to information or knowledge in all intents and purposes has been a race to controlling science (communication), financial and political power. Moreover, intelligence gleaned from users data has not only been profitable, but has also raised long-standing privacy issues.

2019 Big Deals Survey Report: An Updated Mapping of Major Scholarly Publishing Contracts in Europe, European University Association, 2019, states that the consortia including universities, libraries, government representatives, and scientific organisations, representing 30 European countries reported a total annual spend of €726,350,945 on periodicals Big Deals. The proportion of these costs covered by universities is about 72%, or approximately €519,973,578. SPARC Landscape Analysis, SPARC, 2019, observes a company’s revenues per journal article in the region of 4100 USD. Is the staggeringly profitable business of scientific publishing bad for science?, The Guardian, 2017, reports that some publishing and data analytics businesses earn close to 40% profit margins meanwhile scientists get a a bad deal.

Inequality in Knowledge Production: The Integration of Academic Infrastructure by Big Publishers, Posada, 2018, commercial entities traditionally with the publisher role have evolved and increased their control of the research process, publishing process, as well as the research evaluation process.

The Oligopoly of Academic Publishers in the Digital Era, Larivière, 2015, posits that the value added by the publishers did not increase as their rise of ownership in the scientific publishing system over centuries. According to Opening the Black Box of Scholarly Communication Funding: A Public Data Infrastructure for Financial Flows in Academic Publishing, Lawson, 2015, the emergence and growing use of APC model provided a stable revenue stream to publishers, while the technical infrastructure remained relatively unchanged. In Open Access, the Global South and the Politics of Knowledge Production and Circulation, an Open Insights interview with Leslie Chan, 2018, states that adding openness to an asymmetrical and highly unequal system simply amplifies the gap and empowers the already powerful.

A corpus of research articles on the topic of “open” and “free” access to research is collected in Group: Open Access Irony Award, 2019, where individual articles at the time of their publication are themselves inaccessible, require subscription or payment. In Practicing What You Preach: Evaluating Access of Open Access Research, Schultz, 2018, sets to find out how many published research articles about OA fall into the category of publishing their work in paywalled journals and fail to make it open. One explanation of this phenomenon may be that while the research community has shown increasing interest in studying and pursuing open access to research (over several decades, if not more generally over centuries), as well as an overwhelming support for the idea of open access as shown in Highlights from the SOAP project survey. What Scientists Think about Open Access Publishing, Dallmeier-Tiessen, 2011, the actors in the scholarly system are constrained or tied to traditional models of publishing and distribution.

The SPARC Landscape Analysis, SPARC, 2019, reports that some of the major academic publishers are transitioning from providing content to data analytics business. The document provides broad-stroke strategies that higher institutes can consider eg. revising existing data policies, establishing coordination mechanisms, negotiating to ensure institutional ownership of the data and infrastructure and establishing open terms and conditions … re-thinking the institution’s relationship to its data in terms of commercial exploitation, IP ownership, and research investment outcomes.

Vertical Integration in Academic Publishing, Chen, 2019, confirms academic publishing industry’s expansion and control of data analytics by building end-to-end infrastructure that span the entire knowledge production life cycle through mergers and acquisitions. Authors state further vertical integration can potentially have an exclusionary effect on less financially well-endowed journals and institutions, primarily those in the Global South, in their attempt to emulate the western modality of knowledge production.

The article Publisher, be damned! From price gouging to the open road, Harvie, 2014, criticises the large profits made by for-profit publishers while taking advantage of academics’ labours. After internal disagreements (between the publisher and the editorial board), the article was published where the original version introduced a disclaimer warning by the publisher stating that the accuracy of the content should not be relied upon. Such unfortunate turn of events exemplifies how academic freedom can be jeopardised, subject to censorship and manipulation when external parties ultimately control the distribution of knowledge.

In response to paywalled articles with high costs, Sci-Hub is a website that provides free access to research articles and books. The authentication credentials is illicitly obtained authentication credentials to commercial publisher’s online libraries. Sci-Hub has received a variety of responses including praise and criticism from different stakeholders including the research community and the commercial publishers.

In the case of scholarly publishing, standards such as DOI, JIF, and even metadata standards, are not only technical decisions, but are also political decisions with public policy implications. This is because these standards are not neutral, but are designed to privilege certain types of knowledge or outputs, while rendering other invisible.

Some of the observable effects of commercial participants in the scholarly ecosystem on scholarly communication is that they have the means to monitor the usage and interaction of units of scholarly information, censor who or what information gets shared and at what cost, have the agency to control the flow of information, and influence the coupling of their notion of academic impact to a biased incentive system.

The current state of scholarly communication predominantly operates under the notion of requiring or relying on third-party publishers to make scholarly contributions available, as well as setting technical and ethical constraints on what can potentially be a scholarly contribution. Looking at this from another perspective, we can observe that actor-controlled participation is not possible towards “open” participation. I will highlight some of the technical practices (from a plethora of instances) put in place by third-party publishers (irrespective of genuine interest in adhering to the Open Science initiative) which raises a number of user experience (UX) issues. Some commonly observable UX includes:

• Session tracking.
• Requiring cookies to be enabled in order to access the core content.
• JavaScript-driven resources – forcing users to use certain software and enable JavaScript execution.
• Resources (eg. articles, annotations) including tracking software.
• Resources are actively blocked, rate limited, restrictive terms and conditions on reuse for content mining – preventing further research.
• Web crawlers are actively blocked – preventing archiving services to snapshot.
• Proprietary data models and access methods.
• Resources including advertisements.
• Resources including subscription forms to the journal.
• Resources are branded by the publisher.
• Resources change at the discretion of the publisher.
• Persistent identifiers eventually resolving to landing webpages with no deterministic way for machines to discover content.

Researchers exchange units of communication like scientific literature, annotation, datasets, workflows, argumentation, provenance, citation, and software. Knowledge may be represented or presented dynamically, or as a compilation of various independent units (compound). Semantic structure may be embedded in narrative or prose-based scholarly communication as well as used to enhance non-narrative units like experimental results or datasets. Narrative and non-narrative units may in turn include or refer to each other.

While Web resources and their representations can be composed of different kinds of hypermedia, I focus on those that are generally referred to as documents. For instance, articles, annotations, notifications, and profiles when combined can cover a wide range of use cases with respect to units of communication. Each instantiation share common characteristics in that they can be both human- and machine-readable – forming a particular knowledge graph.

Article

An article in the most general sense is a domain-agnostic unit of information encoded in a document. Research contributions including, manuscripts, reports, assessments, technical specifications, news, social media posts and slideshows are some examples for different kinds of articles.

Annotation

Annotations include information that is generally about an association between resources with different intentions. For example, assessing, commenting, liking, bookmarking, describing, linking, are some of the motivations to annotate an article.

Notification

Notifications generally express actor activities. For example, announcements about a scientific article or parts within; a quality assessment of some literature; reports of observations; annotations or social activities.

Profiles

Actors have online profiles where they generally describe themselves, refer to their contacts (address books) or curriculum vitae.

A human-readable format (or medium) is a representation of information that enables humans to read, edit, or act upon. A machine-readable format entails that information can be effectively stored and processed by a machine. Both approaches are equipped with affordances that can be used by respective agents.

Declarative programming paradigm: Declarative programming is a style of building the structure and expressing the semantics of programs, ie. what it should accomplish. The imperative programming style on the other hand is about how the program should execute and change the program state. The declarative approach tends to be short, easy to understand, independent of implementation, less likely to contain errors as well as to easily correct, and tractable. For example, many domain-specific markup languages, such as HTML and XML-family (XSLT, SVG, MathML), and CSS are declarative. HTML simply tells the consuming agent – like a Web browser – what should appear as part of a Webpage. HTML’s readability and ease of interpretation played a role in its adoption – anyone that wanted to express some information in a Webpage can "view source" to learn without any further steps.

RDF as the language: HTML is a prominent media form to publish on the Web. While this is sufficient to cover various use cases in a scholarly information space, it is limited in the sense that the granularity of machine-readable content is based on the classic hypertext model, ie. ultimately a relationship with loose semantics between documents or their parts. It is considered to be limited in terms of capturing domain-specific knowledge. This is in contrast to using RDF as the language to communicate knowledge about arbitrary things at different levels of abstraction. Atomic statements about scientific information and scholarly activities can be expressed, as well as each component of a statement being universally identifiable. The language enables the information to be serialised using different syntaxes eg. HTML+RDFa, Turtle. Perhaps most importantly, the underlying data is intended to be manipulated as a graph of things, where its syntactical representations remaining isomorphic across serialisations.

Human- and machine-readable units: As indicated earlier, the high-level units of communication that I am examining are primarily in the form of prose which may be accompanied with or linked to supplemental (meta)data eg. statistical Linked Data. With this premise, I emphasise on the point that the underlying information is intended for both humans and machines, where each can create and interact with the information through applicable and desired interfaces.

Content negotiation: The HTTP Content Negotiation (RFC 7231) mechanism can be used to serve a possible representation of a resource that the client prefers from a URI. The decision algorithm may be based on different dimensions of content negotiation, eg. media type, character set, encoding, language, time. All things equal, here I consider having simplicity in the design as a desired quality for a server requirement, in order to make data available for the purpose of read-write operations that a client can perform. For instance, to what extent can we serve a resource representation "as is" without having to perform media type conversions in order to satisfy both human and machine consumers? By raising this, it is neither the case that a one-size-fits-all solution is required or desirable.

RDFa: From the available RDF syntaxes, this line of reasoning brings us to encapsulating information using RDFa 1.1 – attribute-level extension – inside host languages like (X)HTML, and various XML-family languages eg. SVG, MathML – language mixing. What makes RDFa in HTML, for instance, an attractive combination is that it maps information patterns in RDF which is ideal for enhanced machine processing, while retaining the same document that is interpretable by humans. Essentially, interoperable information exchange and reuse by machines is intended to work over the RDF graph that is expressed and encoded with RDFa. Here HTML merely acts as the container to encapsulate the underlying information and to offer presentations and interactions (in regardless of the RDF syntax embedded in HTML). RDFa specifies only a syntax and allows independently defined specifications (like vocabularies) to be used. RDFa defines a single, non-domain-specific syntax, used alongside a hosting language’s syntax, so that fragments of information can be consistently interpretable. Existing HTTP servers can virtually serve HTML content without additional server or client-side modules or application logic. Ergo, RDFa in HTML manifests a low barrier to make human- and machine-readable information available from a single URL. The W3C TAG Finding, The Self-Describing Web, 2009, also posits RDFa in HTML as a good practice: To integrate HTML information into the self-describing Semantic Web, use RDFa. Finally, W3C RDFa Use Cases: Scenarios for Embedding RDF in HTML describes the Augmented Browsing for Scientists use case where actors can add RDFa to their articles to indicate the scientific components by reusing the existing vocabularies defined by the scientific community.

Why RDFa: HTML’s extensibility mechanism allows authors to embed data blocks using the <script type=""> mechanism to include content in Turtle, JSON-LD. Information in script are hidden from human view in native HTML interfaces until supplementary processing takes place. For information extraction to take place, the consumer needs to 1) process the HTML, 2) select the script node, 3) parse the content in RDF. On the other hand, RDFa can be included on any HTML tag, and so all human-visible content in HTML can be complemented with machine-readable counterparts in the same context node. This is optimal in avoiding data duplication in the same document which happens to be the case with the other serialisations in RDF, as well as avoiding any further intervention to synchronise data across different nodes. Having a single, unambiguous, and authoritative representation of the information as human-visible and marked as RDFa conforms to the Don’t Repeat Yourself (DRY) principle. It is efficient in that it has no dependency on JavaScript, external or third-party applications to make the hidden machine-readable content be consumable from a human user interface – whereas HTML based user-agents can be expected to conform to user interaction loading Web pages specification. For example, a text-based Web browsers can access and allow interactions with documents. RDFa in HTML is all around a simple but effective design pattern that is able to serve both humans and machines without additional parts or machinery. With all things equal, having all content in HTML is optimal for Web-based archives, as well as helps towards meeting the archiving and awareness functions of scholarly communication.

Content serialisations: Having the canonical representation for articles, annotations, and notifications in HTML+RDFa, still leaves the servers, if so desired, to provide alternative serialisations, eg. Turtle, and JSON-LD, depending on the content negotiation with clients. Articles are represented in HTML+RDFa so that information is usable by both humans and machine consumers while maintaining lowest requirements for publishing, eg. a single URL with full content in HTML+RDFa can be accessible from any HTTP server. No additional requirements are necessary from clients such as JavaScript support or servers with additional RDF based content negotiation.

Separation of concerns: By adopting the progressive enhancement strategy as described in Progressive Enhancement and the Future of Web Design, Champeon and Nick, 2003, for the structural (HTML+RDFa), presentational (CSS), and behavioural (JavaScript) layers, we can allow content and base functionality to be accessible through different media and devices. Ultimately, the unit of communication represented in HTML+RDFa can be accessible – readable – for both humans and machines without requiring any CSS or JavaScript. This approach is considered to cover the base requirement of making the content available with lowest server requirements, ie. practically, today any HTTP server that can serve HTML and a client that can consume HTML.

Affordance: Hypertext has perceivable affordances in that both humans and machines can choose to act. For example, when a Web document is presented with a hyperlink, its users have the option to follow the link, if they so desire, by pressing a key on their keyboard, clicking on it with their pointing-device, or with voice activation and so on. Hovering a link with the cursor can trigger the Web browser to display the full target URL in the status bar of the application, and hence informing the user on what lies ahead before engaging further. If the user is already familiar with the link (visited earlier) or uninterested, they can skip. The interaction with the hypertext is in essence non-linear, ie. we can follow the links as far as we are able to, or interested, and then come back to where we started. Similarly, machines, like Web-crawlers are able to perform exactly the same process. Moreover, hyperlinks marked up with specific relations to their target reference can signal information that can effectively induce unique interactions. In that respect, RDFa in HTML for instance can help enrich and enable the structure for different interaction possibilities, for example though runtime JavaScript or the Web browser’s built-in understanding of the underlying actions. With respect to scholarly referencing, hypertext can not only reference other resources – as typical to print-based expressions – but it can also link.

Units with unique identifiers: Any thing that is deemed to be of importance can be identified on the Web with a URI as per Axiom 0: Universality 1. Having globally unique identifiers for all sorts of objects at fine granularity facilitates precision for interlinked knowledge. We have the means to guide both human and machine users to better discover and exchange scientific and scholarly objects. Thereby making it possible to fulfill the registration function for virtually anything.

Markup patterns: HTML has several extensibility mechanisms, eg. class, data-*, rel attributes, and meta, script elements to support vendor-neutral markup patterns. In addition to HTML’s common infrastructure, it is possible for documents to signal formal grammars, eg. HTML profiles and XML schemas, to announce the structure of documents such that the machine-processing instructions are well-defined; predetermined and fixed. In practice, grammars are used to help extract markup patterns according to specified definitions. Although profiles are no longer supported in HTML5, DroppedAttributeProfile, W3C, HTML Working Group, 2010, they were originally intended to assign additional meaning that would otherwise remain local (private) to the document. For instance, Gleaning Resource Descriptions from Dialects of Languages (GRDDL) provided a way for custom XSL Transformations (XSLT) to handle out-of-band transformation workflows, and to interpret the structure and semantics in the original document.

Formal grammars: While HTML markup with formal grammars has the quality of being efficient for cooperating systems, it comes at the cost of wider interoperability. That is, if an application is designed to only work with a particular flavour or subset of HTML, then it would not be able to work with arbitrary HTML documents on the Web. Hence, having profiles or grammars in and itself not an issue, however its effects are likely to be that they are less interoperable than they can be. This is in contrast to applications safely ignoring information patterns they do not understand or interested in using, even if they have their preferred internal grammars or HTML templates. From the consuming side, if the application’s information reuse is dependent on HTML a specific schema or tree structure, effectively locks the application into coping with those patterns, and ultimately less flexible about handling expressions that the application is unfamiliar with. The effect is a constrained or a closed information system.

Here I describe a wide range of well-known Semantic Web vocabularies that can be used to model and describe units of communication. It is not intended to be an exhaustive list but to provide an overview on the kind of things that could be potentially described, anywhere from publishing articles, scholarly activities, to datasets holding scientific measurements.

General purpose: DCMI Metadata Terms is used to describe digital and physical resources along with metadata on provenance and rights. schema.org is composed of a common set of schemas that occurs in Web pages. These vocabularies are widely used on the Web and are generally domain agnostic, making them suitable to describe document-level concepts with respect to scholarly units.

Publishing and referencing: Bibliographic Ontology Specification and Semantic Publishing and Referencing Ontologies (SPAR) cover a wide range of concepts in the publishing domain while being agnostic about the types of content, eg. document description, bibliographic resource identifiers, types of citations and related contexts, bibliographic references, document parts and status, agents’ roles and contributions, bibliometric data and workflow processes. The SPAR Ontologies, Peroni, 2018, describes ontologies for bibliographic resources and their parts (FaBiO, FRBR-DL, DoCO, DEO, DataCite); citations of scholarly resources (CiTO, BiRO, C4O); publishing workflow (PRO, PSO, PWO, SCoRO, FRAPO); metrics and statistics for bibliographic resources (BiDO, FiveStars).

Knowledge Organisation: Simple Knowledge Organization System (SKOS) defines a common data model for structures like thesauri, taxonomies, classification schemes and subject heading systems in library and information sciences. SKOS is suitable to bridge different scientific communities to organise concepts by providing definitions, links, and possible mappings across collections.

Participation: Actors are integral to social Web activities and the four functions of scholarly communication. Generally speaking, agent – human or machine – profiles can be described with the Friend of a Friend (FOAF) vocabulary to have a combination of what or who they are, what they have created, who they know, their memberships, and so forth. vCard Ontology - for describing People and Organizations describes the mapping of the vCard specification to RDF/OWL in order to allow compatible representations for personal data interchange. SIOC Core Ontology Specification provides concepts to describe information from online communities like weblogs, message boards, wikis, etc., as well as connections between their content items. Activity Vocabulary provides specific activity structures and types for objects and links (eg. article, event, place, mention), actors (eg. application, group, person, service) and for past, current or future social activities (eg. announce, create, like, invite, question). While the core vocabulary covers a wide range of activities, it can be extended to cover domain-specific scientific activities. The Web Annotation Vocabulary underpins the annotation model and is used to express information about a set of connected resources, typically conveying some sort of a relationship where the body of a resource is about the target resource. Embedding Web Annotations in HTML describes and illustrates potential approaches for including annotations within HTML documents using current specifications like RDFa or JSON-LD. Selectors and States describes the use of annotation Selectors as URI fragment identifiers relying on the formal specification and the semantics in the data model. Any resources can advertise a location where it can receive Linked Data Notifications for social activities eg. invitation to review an article, annotation being created. I describe LDN in detail later serving as one of the foundational components towards a decentralised and social Web. In order to indicate an agent’s cognitive pattern within contexts, their temporal dynamics and their origins, The Cognitive Characteristics Ontology can be used eg. a human language competence.

Rights and Responsibilities: Creative Commons Rights Expression Language lets licenses to be described – jurisdiction, permissions, requirements, license and work properties – and having them attached to digital works. The Creative Commons licenses is intended to enable sharing and reuse of creative works and knowledge, by specifying how something – like scientific and scholarly resources – may be copied, distributed, edited, remixed, and built upon. The Open Digital Rights Language (ODRL) makes it possible to express permissions and obligations about the usage of content and services. ODRL Vocabulary & Expression describes how to encode such policies.

Access Control and Certificates: The Web Access Control (WAC) mechanism with the Access Control List (ACL) vocabulary describes authorization policies in particular to access and operations – read, write, append, control – that can be done on Web resources by agents or groups. The Cert Ontology 1.0 can be used to indicate digital certificate information for agents.

Provenance: The PROV Ontology (PROV-O) can be used to represent and interchange provenance information generated in different systems and under different contexts. It can be also be specialized for modeling application-specific provenance details in a variety of domains, like for instance The OPMW-PROV Ontology to describe workflow traces and their templates, which also extends the P-Plan Ontology that is designed to represent scientific processes. The Wf4Ever Research Object Model is used to describe workflow-centric Research Objects: aggregations of resources along with annotations on those resources relating to scientific workflows. Verifiable Claims Data Model 1.0 is aimed at expressing a claim; statement made by an entity about a subject, on the Web in a way that is cryptographically secure, privacy respecting, and automatically verifiable. The Memento terms vocabulary contains a set of terms for the Memento Framework to facilitate obtaining representation of resource states. Link Relation Types for Simple Version Navigation between Web Resources (RFC 5829) defines link relation types to related to navigating between versioned Web resources, such as latest and predecessor versions and working copies.

Design Lifecycle: Design Intent Ontology can be used to to capture the knowledge generated during various phases of the overall design lifecycle like for instance the artifacts about software or technical specification requirements, issues, solutions, justifications and evidence.

Datasets: For multi-dimensional statistical data, the RDF Data Cube vocabulary can be used to both model the structure of the hypercube as well as aggregate data which uses its structure definition. The structure is typically an arbitrary number of components: dimensions that helps to identify an observation, and one or more measurements about the phenomenon being observed. The DDI-RDF Discovery Vocabulary (Disco) enables publishing and discovery of metadata about datasets like research and survey data. Disco is generally concerned with microdata descriptions that is on a lower level of aggregation than the RDF Data Cube. It is not intended to represent the data itself, but only its structure, where the record-level raw data in its original format is only referenced. Disco also enables a way to describe the aggregation methods that was used to collect the data. The Semantic Sensor Network Ontology (SSN) can be used for describing sensors and their observations, the involved procedures, the studied features of interest, the samples used to do so, and the observed properties, as well as actuators. Describing Linked Datasets with the VoID Vocabulary (VoID) is concerned with metadata about RDF datasets to help with deployment, discovery of data and services, cataloguing and archiving of RDF datasets, as well as linksets between datasets. Evaluation and Report Language (EARL) can be used to describe the test results and facilitate their exchange between applications. It provides reusable terms for generic quality assurance and validation purposes.

Scientific expressions: Semanticscience Integrated Ontology can be used to describe scientific experiments, for example including their procedure, hypothesis, objectives, study design, analysis, and observations. STATO: the statistical methods ontology covers statistical methods, tests, conditions of application, results, as well as aspects of experimental design and descriptions. Nanopublication Guidelines helps to publish scientific assertions that can be uniquely identified with associated context, attributed to their author, as well as help to preserve associated provenance. Micropublications can be used to formalise the arguments and evidence in scientific publications.

Software Projects: Description of a Project (DOAP) can be used to describe (open source) software projects like their repositories, bug databases, maintainers, technical specifications they implement, programming languages they use or the platform they run on.

We acknowledge the diversity of people using the Web, not only the actors in scholarly communication, but anyone that may create or reuse information. Our aim is to have inclusive designs for wide range of people and their abilities. I outline and borrow key initiatives and solutions on content accessibility, accessible applications, authoring tools, and internationalisation. I refer to accessibility (a11y) in the widest sense that any agent; human, machine, or other can effectively access information and participate.

Web Content Accessibility: The accessibility of units of communications can also be seen as an aim for all-inclusive design that is usable by humans with widest possible range of abilities and situations. For this, I defer to W3C’s Web Content Accessibility Guidelines (WCAG) to cover an array of recommendations to make content accessible to a wider range of people regardless of any disability, limitation, or sensitivity, through different media and interfaces. The goal is to provide a reliably mutually consistent expression of the content. For human-centric scholarly communication to thrive, units of communication should be targeted to aim to meet highest level of conformance criteria. W3C WCAG 2.1 provides a range of guidelines that can be adopted for better content accessibility:

Perceivable

Information and user interface components must be presentable to users in ways they can perceive.

Operable

User interface components and navigation must be operable.

Understandable

Information and the operation of user interface must be understandable.

Robust

Content must be robust enough that it can be interpreted by a wide variety of user agents, including assistive technologies.

Accessible Applications: The W3C Accessible Rich Internet Applications (WAI-ARIA) recommendation provides an ontology to help assistive technologies to provide a consistent user interface and understanding of the objects. Host languages like HTML and SVG can include the ARIA ontology: roles to alert the purpose of an element; properties to indicate an elements relationship to other things; states to indicate what the element is doing, as well as alerting users about changes in state. ARIA helps to inform consuming assistive technologies such as screen magnifiers, screen readers, text-to-speech software, speech recognition software, alternate input technologies, and alternate pointing devices to create a particular accessibility tree, and to adapt the user interface to a form that works for the person, eg. a screen-reader can read the menu items, or selected option is receiving keyboard focus. WAI-ARIA is extensible and has a number of accessible API mappings. Digital Publishing WAI-ARIA Module 1.0 (DPUB-ARIA) specialises WAI-ARIA’s ontology to enable semantic navigation, styling and interactive features in context of digital publishing. WAI-ARIA Graphics Module is another extension of WAI-ARIA aimed to support structured graphics such as charts, graphs, technical drawings and scientific diagrams, to assistive technologies in order improve accessibility of graphics or diagrams through detailed annotations.

Authoring Tool Accessibility: In order to contribute to the proliferation of Web content that is accessible to a broad range of people, authoring tools would need to be accessible as well. The W3C Authoring Tool Accessibility Guidelines (ATAG) is a recommendation to assist with the design of authoring tools. The guidelines have a success criteria covering two areas:

• Make the authoring tool user interface accessible
• Support the production of accessible content

User Agent Accessibility Guidelines: To support the general principles for the development of accessible user agents, ie. any software that retrieves, renders and facilitates end-user interaction with web content, the W3C User Agent Accessibility Guidelines (UUAG) specifies three layers of guidance for developers to integrate:

• overall principles: perceivable, operable, understandable, programmatic access, specifications and conventions;
• general guidelines to provide a framework to make user agents more accessible to users with disabilities;
• and success criteria in order to test conformance

Internationalization and Localization: Adaptability of content and software to the needs of target audiences helps towards accessibility. For example, the mechanisms to cater information and interfaces so that people from any culture, region, or language preference can participate better. The W3C internationalization (i18n) and localization (l10n) initiatives and best practices helps towards this end. Internationalization refers to the design and development of mechanisms so that adaptable localization can take place in users’ environment.

Internationalization of Web content and technologies is intended to make it possible for people to use them with different languages, scripts, and cultures. For example, content authors can:

• include links to navigate to different languages of the content;
• declare the base language of a document, indicate multiple languages and their directional flow – to help with translations;
• use Unicode character encoding, eg. UTF-8, in data forms and text to ensure correct effects;
• check and minimise inappropriate cultural bias, and improve translatability;
• restrict markup use to structure and semantics.

The localization of content would mean that user’s preferred (or acceptable) visual design can be presented. For instance, human-readable numeric, date and time formats can be adapted to what we are familiar with; symbols, icons, and colours can be anywhere from what’s culturally acceptable, familiar or comfortable for us to use; text and graphics can be normalised or transformed to minimise misinterpretation.

Archivability of Web resources refers to the ease with which the content, structure, functionality, and front-end presentation(s) of a website can be preserved and later re-presented, using contemporary web archiving tools.

In CLEAR: a credible method to evaluate website archivability, Banos, 2013 posit that Web archivability can be measured by five facets: accessibility as network access to content; standards compliance in terms of information using common open formats and specifications; cohesion as information being independent of external support; the level of metadata that is available alongside the content; and, server’s performance. Such facets can be used to quantify website archivability.

In the Web development community, a rule of thumb to improve archivability of Web content is to aim for general standards compliance and content accessibility. In the case of standards compliance, well-formed and valid markup ensures internal integrity of Web documents. As for better accessibility, incorporating WCAG where applicable can help user-agents, including text-only Web browsers, Web crawlers, and other agents with minimal capabilities to parse and render content. With respect to traditional Web documents, the information that is visible, observable, or human-readable from HTML is considered to set the lowest barrier to obtain content. As CSS and JavaScript are concerned with presentation and behavioural layers respectively, they play a secondary role towards archiving of "content".

The impact of JavaScript on archivability, Brunelle, 2015, study the quality of archived Web resources (based on URIs from Twitter and Archive-It) and report that JavaScript-driven content played a significant role in the reduction of preservation and recall of content. For example, resources referring to external scripts may load unsuccessfully given that their own independent availability. Authors refer to the notion of deferred representations where the final state of a resource is when all code and events have finished executing. Hence, representations rely on JavaScript to fully render do not serve well for archival crawlers that are unequipped to handle JavaScript. On the other hand, certain archival services include a feature known as headless browser that is capable of processing the resource similar to common GUI-based Web browsers with JavaScript enabled, and thus capturing the final state of the representation once all scripts are executed. This too however may be problematic for content that dynamically changes based on the instructions of the script. For example, a script that presents temporal content may be different at each time the archived resource is recalled.

While evaluation services like ArchiveReady check for website archivability based on the five metrics, it is not able to determine the accuracy of intended information. As the content in the HTML representation and the state of rendered content (after script execution) may differ, their delta can be compared to what is intended for archiving. In Significance is in the Eye of the Stakeholder, Dappert, 2009, contend that a stakeholder can specify the characteristics of resource representations and the conditions for preservation. For human-readable documents, it can be reasoned that the rendered content is what is intended for archiving by the publisher.

As statistical data is inherently highly structured and comes with rich metadata (in form of code lists, data cubes etc.), the application of the Linked Data design patterns is inherently suitable for representation and reuse on the Web. There exists no simple, standardised or (semi)automated way to transform statistical data into Linked Data since the raw data comes in different shapes and forms.

While access to statistical data in the public sector has increased in recent years, a range of technical challenges makes it difficult for data consumers to take advantage at ease. These are particularly related to the following two areas:

• Automation of data transformation of data from high profile statistical organizations.
• Minimisation of third-party interpretation of the source data and metadata and lossless transformations.

Statistical Data and Metadata eXchange (SDMX) is an ISO standard which provides the possibility to consistently carry out data flows between publishers and consumers. SDMX-ML (using XML syntax) is considered to be the gold standard for expressing statistical data. It has a highly structured mechanism to represent statistical observations, classifications, and data structures. Given that SDMX is arguably the most widely used standard for statistical data exchange (among statistical agencies and others), a great amount of statistical data about our societies is yet to be discoverable and identifiable through the Open Web Platform. Development teams often face low-level repetitive data management tasks to deal with someone else’s data. Within the context of Linked Data, one aspect is to transform this raw statistical data eg. SDMX-ML, into an RDF representation in order to be able to use publicly available data in a uniform way.

Linked SDMX Data, Capadisli, 2013, ie. my colleagues and I, contribute as follows:

• an approach for transforming SDMX-ML based on XSLT 2.0 templates and the implementation which transforms SDMX-ML data to RDF/XML
• The extract, transform, load ETL process: a) retrieval of SDMX-ML data from APIs, b) their transformations to RDF/XML, and c) enrichment, d) storage in an RDF data store, and e) publicly accessible Linked Data publication based on the statistical data from the following agencies:
  • Australian Bureau of Statistics (ABS)
  • Swiss Federal Statistical Office (BFS)
  • Bank for International Settlements (BIS)
  • European Central Bank (ECB)
  • Food and Agriculture Organization of the United Nations (FAO)
  • Federal Reserve Board (FRB)
  • International Monetary Fund (IMF)
  • Organisation for Economic Co-operation and Development (OECD)
  • UNESCO Institute for Statistics (UIS)

Each Linked Statistical Data (LSD) endpoint:

• modeled using RDF, RDFS, XSD, OWL, XSD, DC Terms for general purpose descriptions; the RDF Data Cube vocabulary to describe multi-dimensional statistical data; PROV-O is used for provenance coverage; SKOS and XKOS to cover concepts, concept schemes and their relationships to one another;
• cross dataset and concept scheme interlinking;
• uniquely identifiable provenance level information at retrieval, transformation, and post-processing phases;
• licensed under CC0 1.0 Universal Public Domain Dedication;
• follows a URI template, where the URIs are versioned with respect to the original data counterparts
• data dumps;
• queryable SPARQL endpoint;

The results are part of the 270a.info LSD Cloud. The LSD Cloud is also part of the broader Linked Open Data Cloud (LOD) Cloud.

What the LSD Cloud enables is that highly structured and interlinked statistical data and classifications that can be queried and used by decentralised applications.

In Linked Statistical Data Analysis, Capadisli, 2013, puts forward an approach based on decentralised (and federated) structured queries to retrieve statistical data from various SPARQL endpoints, conducting various data analyses (eg. regression analysis), and providing the results of the analysis as Linked Data back to the user. The system demonstrating the mechanism stores the analysis in RDF to enable future discovery and reuse eg. researchers looking up statistically significant results based on a set of indicators for a specific reference area. As a result, distributed linked statistics with accompanying provenance data can be more easily explored and analysed by interested parties.

The approach expects that the data is modelled using the RDF Data Cube vocabulary and has Well-formed cubes. Essential checks for integrity constraints include: 1) a unique data structure definition (DSD) is used for a dataset, 2) the DSD includes a measure (value of each observation), 3) concept dimensions have code lists, and 4) codes are from the code lists.

In order to compare variables in observations across statistical datasets, there needs to be an agreement on the concepts that are being matched for in respective observations. In the case of regression analysis, the primary concern is about reference areas (ie. locations), and making sure that the comparison made for the observations from dataset_x (independent variable) and dataset_y (dependent variable) are using concepts that are interlinked (using the property skos:exactMatch). Practically, a concept eg. Switzerland, from at least one of the dataset’s code lists should have a relation to the other dataset’s concept. It ensures that there is a reliable degree of confidence that the particular concept is interchangeable or the degree in which the concepts being comparable. Hence, the measure corresponding to the phenomenon being observed, is about the same location in both datasets. To this end, concepts in the datasets were interlinked.

In order to foster trust and confidence for data consumers (human and machine), the analysis is accompanied with provenance data. The analysis includes a Oh yeah? reference (in HTML as well as in RDF serializations) intended to guide the data consumer to a resource about the provenance activity – using the PROV-O vocabulary – about the performed analysis. These previously generated provenance activities provide links to all data sources which were used for the analysis, query construct for data aggregation, as well as metadata about the used tools, assigned license, production timestamps, and responsible agents for the generated analysis. Thus, in addition to analysis metadata, the user is able to track the data all the way back to its origins (at the statistical agencies), and reproduce or compare their results.

The proposed approach focuses on the scalability of the system with minimal human intervention. That is, new statistical dataset can be independently published while being expressed with relevant statistical vocabularies, and sufficient interlinks between concepts, then applications can discover them and perform queries in a uniform way. Hence, applications only need to be made aware of the location of new datasets.

Building on the notion of modularity, we consider that interoperability occurs between two classes of interactions, senders and receivers, or between consumers and receivers, when the sender/consumer makes a request to the receiver, and the receiver sends the expected response.

The data formats which are used by a notification specification must be chosen to maximise interoperability between implementations. Structure of Scholarly Information makes the argument for using RDF as the language to express human- and machine-readable information on the Web. Allowing LDN actors – senders, receivers, and consumers – to handle data irrespective of the particular RDF serialisation permits flexibility, however it can also be costly to support. We take into account:

• Application interoperability eg. should applications support all current RDF formats, as well as new versions or formats in the future?
• Maintenance of RDF parsers and serialisation libraries eg. will the RDF libraries that applications use be kept up to date?
• Complexity of their inclusion in applications eg. what are the costs or concerns of using all RDF formats?
• Run-time efficiency eg. how well will applications perform going forward?

Why JSON-LD: To address these issues, choosing a single RDF serialisation to require is necessary for consistent interoperability, as well as keeping processing requirements or external code dependencies minimal. While any RDF notation would satisfy that need, we decided that LDN requires all applications to create and understand the JSON-LD syntax, both for the contents of Inbox as well as for individual notifications.

To some extent, JSON-LD is compatible with existing JSON libraries or in some cases native programming language data structures. For example, Web browsers have a built-in interpreter for JavaScript that can handle JSON objects, that can take advantage of native object operations. Same data processing and manipulation operations are also available in a Node.js environment. Having said that, JSON-LD is a format that serves to be an RDF notation, and can also be processed as plain JSON. Hence, JSON-LD is just one convenience to certain kinds of environments that the application is confined to. JSON-LD is also advantageous in being familiar for developers who are used to JSON-based APIs but not RDF.

Content negotiation: Optionally, applications may attempt to exchange different RDF serialisations by performing content negotiation. As in compliance with HTTP principles, a sender may make an HTTP OPTIONS request to the receiver to determine the RDF content types accepted by the server. A receiver can expose Accept-Post headers for senders, and so the sender can serialize the notification in the request body accordingly. Consumers can send Accept headers to receivers to signal which RDF content types they prefer and capable of processing.

By mandating one RDF syntax (JSON-LD) for applications to communicate with, as well as permitting other syntaxes to discover and exchange notifications, compatibility across applications is assured.

Target ownership: As per AWWW’s URI Ownership, the Inbox is ultimately controlled by the owner of LDN Receiver, and subject to third-party access to HTTP headers and content. Hence, publishers of the resources advertising an Inbox (target) are expected to be aware of using servers they trust or control for the Inbox location.

Constraints: One way to filter unwarranted notifications from being created on the server and exposed is where Inbox URLs announce their own constraints eg. Shapes Constraint Language (SHACL), ShEx, Web Annotation Protocol, via an HTTP Link header or body of the resource with a rel value of ldp:constrainedBy. For example, an LDN receiver may want to allow notifications with a certain shape; a model specifying allowed and required values in a notification. This is so that senders can comply with the advertised constraint specification or the receiver may reject their notification. Rejecting notifications which do not match a specific pattern in their contents, or the shape of the data, is one way to filter. For example, if the Inbox owner knows that they will only ever use a consuming application which processes friend requests, they can configure their receiver to filter out anything that does not match the pattern for a friend request, helping their consumer to be more efficient. If the notification constraints are also advertised by the receiving service as structured descriptions, generation and consumption of the notifications can be further automated. Constraints are particularly useful towards providing notifications that meets a certain criteria so that it can be used effectively by consuming applications.

Authenticated inboxes: The requirement for reusable notifications could be seen as a potential risk for privacy. The LDN Receivers are expected to consider implementing access control on the Inbox URL as well as the individual notification URL and may restrict reading and writing to a whitelist of trusted senders. Various authentication methods which could be used alongside LDN with the idea that notifications are not necessarily public but only visible or reusable by intended parties. The receivers decide (based on their use case) which consumers (based on any criteria) can reuse. They achieve this by setting authentication and authorization settings on the notifications. As different authentication mechanisms are appropriate for different applications, the notification protocol should ideally be usable alongside various methods such as clientside certificates, eg. WebID+TLS, token-based, eg. OAuth 2.0, or digital signatures.

Personally identifiable information: Notification payloads may contain any data including that which identifies the sender or the receiver. As access to the Inbox and notification data is under the control of the receiver, they ultimately determine what information may be exposed to the world. Once fetched by a third-party, any piece of information could potentially be subject to further (unauthorised) distribution and reuse. Hence, this is orthogonal to whether the notifications are ephemeral or persistent.

Security and Privacy Review: The LDN specification includes a Security and Privacy Review covering threat models categorised as: Passive Network Attackers, Active Network Attackers, Same-Origin Policy Violations, Third-Party Tracking as outlined in W3C Self-Review Questionnaire: Security and Privacy.

This section demonstrates some interactions between a server with a target resource, and among sender, receiver, and consumer implementations.

Discovering an inbox:

GET / HTTP/1.1
Host: csarven.ca
Accept: application/ld+json

HTTP/1.1 200 OK
Content-Type: application/ld+json

{
  "@context": "http://www.w3.org/ns/ldp",
  "@id": "https://csarven.ca/#i",
  "inbox": "https://csarven.ca/inbox/"
}

Consumer requesting the inbox and receiver’s response:

GET /inbox/ HTTP/1.1
Host: example.org
Accept: application/ld+json
Accept-Language: en-GB,en;q=0.8, en-US;q=0.6

HTTP/1.1 200 OK
Content-Type: application/ld+json
Content-Language: en

{
  "@context": "http://www.w3.org/ns/ldp",
  "@id": "http://example.org/inbox/",
  "contains": [
    "http://example.org/inbox/5c6ca040",
    "http://example.org/inbox/92d72f00"
  ]
}

Consumer getting a notification:

GET /inbox/14a792f0 HTTP/1.1
Host: example.org
Accept: application/ld+json, text/turtle, application/xhtml+xml, text/html
Accept-Language: en-GB,en;q=0.8, en-US;q=0.6

HTTP/1.1 200 OK
Content-Type: application/ld+json;profile="https://www.w3.org/ns/activitystreams"
Content-Language: en

{
  "@context": [
    "https://www.w3.org/ns/activitystreams",
    { "@language": "en" }
  ],
  "@id": "http://example.org/inbox/14a792f0",
  "@type": "Announce",
  "actor": {
    "@id": "https://csarven.ca/#i",
    "name": "Sarven Capadisli"
  },
  "object": {
    "@context": "http://www.w3.org/ns/anno.jsonld",
    "@id": "http://example.net/note",
    "@type": "Annotation",
    "motivation": "http://www.w3.org/ns/oa#assessing",
    "rights": "http://creativecommons.org/licenses/by/4.0/"
  },
  "target": "http://example.org/article",
  "updated": {
    "@type": "http://www.w3.org/2001/XMLSchema#dateTime",
    "@value": "2016-06-28T19:56:20.114Z"
  }
}

Sending a notification:

POST /inbox/ HTTP/1.1
Host: example.org
Content-Type: application/ld+json;profile="https://www.w3.org/ns/activitystreams"
Content-Language: en

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "@id": "",
  "@type": "Announce",
  "actor": "https://rhiaro.co.uk/#me",
  "object": "http://example.net/note",
  "target": "http://example.org/article",
  "updated": "2016-06-28T19:56:20.114Z"
}

Response to send request:

HTTP/1.1 201 Created
Location: http://example.org/inbox/5c6ca040

Sender initiates content-negotiation and sends notifications:

OPTIONS /inbox/ HTTP/1.1
Host: example.org

HTTP/1.1 200 OK
Allow: GET, HEAD, OPTIONS, POST
Accept-Post: application/ld+json, text/turtle

POST /inbox/ HTTP/1.1
Host: example.org
Content-Type: text/turtle
Content-Language: en

@prefix as: <https://www.w3.org/ns/activitystreams#> .
@prefix cito: <http://purl.org/spar/cito/> .
<> a as:Announce
  as:object <https://linkedresearch.org/resources#r-903b83> ;
  as:target <https://csarven.ca/dokieli-rww#architecture-and-technologies> .
<https://linkedresearch.org/resources#r-903b83>
  cito:citesAsPotentialReading
  <https://csarven.ca/linked-data-notifications#protocol> .

Derived From

https://linkedresearch.org/ldn/tests/summary#ldn-report-receiver

Derived On

2019-01-05

The receiver implementations are either LDP-based, extension of LDP, generally LD-based platforms, stand-alone libraries, or integrated into existing domain specific systems, eg. personal websites. For instance, implementations range from minimum proof of concepts like DIY Inbox to enterprise-ready platforms like Apache Marmotta or Virtuoso Universal Server. Sloph, The Presentation of Self on a Decentralised Web, Guy, 2017, and IndieAnndroid are part of personal publishing and quantified self platforms geared around social media-like interactions. The Scholastic Commentaries and Texts Archive (SCTA) inbox is used for annotating scholarly manuscripts, and serving notifications about the annotations. ldn-streams is a specialised receiver that is capable of accepting and serving notifications as RDF streams. distbin is a general purpose “pastebin” to store data. The rest of the implementations are non-domain specific servers that can be used for different purposes.

Derived From

https://linkedresearch.org/ldn/tests/summary#ldn-report-consumer

Derived On

2019-01-05

The consumer implementations so far include streaming-capable implementations, stand-alone libraries, and domain specific applications. As part of the International Image Interoperability Framework (IIIF) Discovery Support Technical Specifications Group, an LDN implementation aggregates metadata to discover resources made available by a IIIF service. Later in this thesis, I describe how LDN is implemented in context of dokieli, a clientside authoring and publishing application that can consume notifications to discover independently published annotations on the Web.

Derived From

https://linkedresearch.org/ldn/tests/summary#ldn-report-sender

Derived On

2019-01-05

Some of the sender implementations include streaming-capable implementations of LDN, stand-alone libraries, and domain specific applications. Some of the sender implementations also fulfill the role of a consumer. For instance, Linked Edit Rules checks the consistency of statistical datasets against structured constraints, and delivers the consistency report as a notification to the user. py-ldnlib is a library that can be reused by other applications to send LDN notifications.

This section covers a direct comparison between existing notification mechanisms from the Overview of Web Notification Systems with Linked Data Notifications, followed by a discussion of the trade-offs that were necessary to realize certain benefits of LDN. The comparison criteria include our design considerations (Rx) along with additional technical information which helps to capture some design differences (Tx).

The figure Comparison of notification mechanisms is derived from https://csarven.ca/linked-data-notifications#comparison-of-notification-mechanisms and modified to include WebSub.

All of the notification mechanisms specify multiple roles, with the Sender (S) being common across all of them. LDN is the only mechanism which describes the Consumer (C) role; this explicitly highlights how notifications can be useful to applications other than those concerned with sending and receiving. For mechanisms which include a Subscriber (U) role, the Receiver (R) is essentially also the Consumer, so Receivers are less generic and require specialised payloads. In this respect, LDN is a more modular specification. LDN Receivers need not be concerned with additional actions beyond storing and exposing notifications; complex and domain-specific uses of the notifications are left to Consumers, resulting in a lot more flexibility in terms of lightweight end-user applications.

WebSub differs in terms of the number of roles, but not the actual functions of the specified roles. The hub is an intermediate party which plays the role of a Receiver and a Sender, relaying notifications from the original Sender (or publisher) to the Subscriber (or ultimate Receiver), but nonetheless only makes notifications available to the Subscriber and no other applications.

LDN enforces that notifications should be reusable, and specifies how Receivers should make notifications available for Consumers. The other mechanisms either consider notification reuse as inapplicable or leave it optional and unspecified.

This is also tied to to persistence and retrievability in that LDN requires notifications to be identified by dereferenceable HTTP URIs. The other notification mechanisms treat notifications as ephemeral resources which only exist until they are delivered to a Receiver.

The mechanisms with the fat ping approach require structured data in XML or JSON-LD. Only LDN requires the use of RDF through JSON-LD, and optional content negotiation for other RDF syntaxes. Thus, notification payloads can include references to other retrievable resources, or can embed all relevant data directly in the payload.

Approaches which send only URLs (as with linkback) rely on the receiver interpreting a third-party resource, which may or may not contain structured content or be under the control of the sender. Approaches which offer additional guidance to aid the receiver in interpreting the source document(s) nonetheless still restrict the sender. LDN therefore offers flexibility to senders, increasing the potential uses for the notification mechanism. LDN compensates for increased complexity on the receiver’s end by recommending filtering mechanisms, and moving most of the burden of understanding notification contents to the consumer role. In placing no constraints on the notification payload, LDN enables a sender to be precise and lossless with the data it is transmitting. As such LDN can cover a broader variety of use cases.

LDN recommends that Receiver applications perform some processing of the payload. This is to enable users to configure their Receivers to filter out spam or customise them for particular domains, but is not required; such filtering could also be left to or supplemented by Consumers. In contrast, Semantic Pingback and Webmention require receivers to parse the form-encoded payload and then to perform fetching of the URLs within and additional parsing of the documents those URLs identify. This is required in order to verify the message (by finding the target URL). Such processing in LDN is left at the discretion of the Receiver; linkback style verification is not required, but also not forbidden.

As WebSub requires subscribing as part of its core design, hubs are expected to perform specific actions on receipt of a subscription request (where the payload is tightly constrained), but not necessarily to process the contents of a notification sent by a publisher (where the payload can be anything; how they handle this depends on the content-type).

Though LDN does not specify a subscription mechanism in the same vein as WebSub, thanks to the modularity of specification conformance classes and the flexible nature of the notification contents, it is possible for applications to perform subscription-like interaction by taking on multiple roles.

LDN requires that Senders and Consumers are equipped to discover Inboxes (receiving endpoints) through both HTTP headers and the body of an RDF resource. This slightly increased complexity around Inbox discovery for Senders and Consumers is a worthwhile tradeoff to lower the bar for publishers of target resources; Inboxes can be advertised from any document on the Web (through HTTP headers), and attached to even non-informational resources (through RDF statements). This increases the potential use cases for LDN.

It is difficult to objectively measure performance metrics of different mechanisms as the various implementations use different servers and programming languages whose efficiency do not necessarily have any bearing on the protocols themselves. Mechanisms which require large payloads may increase the time for HTTP requests to be processed, and mechanisms that require verification add an additional network request to complete the process.

There are several technical specifications that LDN may be interoperably combined. Here I discuss a few.

Relationship with Linked Data Platform: An LDN Receiver is not dependent on a complete implementation of the LDP specification, but comprises an easy-to-implement subset. An LDN Inbox is comparable to an LDP BasicContainer. LDP does not define how clients discover LDP Containers. Through LDN’s inbox discovery, LDP Containers for the purpose of holding notifications can be discovered. Thus, the core features necessary to exchange notifications between LDN applications is effectively possible with existing, as well as future implementations of LDP.

Relationship with ActivityPub: ActivityPub uses LDN’s targeting and delivery mechanism with some specific constraints. Notifications must use a single AS2 Activity in compact JSON-LD syntax. Receivers are required to authenticate requests made by Senders, as well as verify the existence of an object (that the activity is about) that is mentioned in the notification by fetching its source from the origin server. It is possible for LDN Senders to deliver notifications to AP servers. It is also possible for AP clients to deliver messages to LDN Receivers with some bridging.

Relationship with Fedora API: The Fedora API Specification is in the process of being formalised (as an extension of LDP) by the Fedora community. As LDN can be used to support external integrations, Fedora API’s repository event stream draws upon the LDN specification, allowing LDN consumers and senders to react asynchronously to repository events. Fedora implementations were included in the LDN implementation reports.

Linked Data vocabularies: As LDN is agnostic about the contents of a notification, any Linked Data vocabulary can be used, including the Vocabularies I have previously discussed in Structure of Scholarly Information.

Relationship with other federation protocols: Further similarities and differences between LDN and other federation protocols and social APIs are described in W3C Note Social Web Protocols. For example, a bridging code can be applied by applications implementing different specifications to achieve further interoperability between systems.

At the beginning of this section, four requirements relating to Degree of Control were listed. LDN addresses each of these as follows:

Actors can use their preferred applications to discover, reuse, and send notifications

LDN was developed as a W3C standard, with strict criteria for interoperability. Applications which implement any of the roles of the protocol can confirm their conformance with the test suite, and are thus expected to be able to interoperate with other conformant applications. Actors should be able to choose any available conformant application and maintain expected notification-related functionality.

Users store incoming notifications where they prefer

Any conforming LDN Receiver (server) under a particular user’s control – which they trust, and are authorised to read – can be chosen to store notifications.

Users can switch between applications without having to move their data

LDN specifies the mechanism for information exchange, so it is possible to switch between Receiver applications at the protocol level without any further action (implementation-specific details like database storage mechanism may require additional work for portability of data already stored there; this is out of scope of the LDN specification). A user’s Receiver, where notification data is stored, has no bearing on their ability to alternate between sending and consuming applications as needed.

Users change the location of their data without having to change their application

Users can serve their inbox and the notifications from different HTTP URIs provided that they do not mind persistence. In this case, the target resource’s inbox location would simply point to the new location. As inbox discovery starts from looking up a target resource’s description, the new inbox location will be discovered as before.

The LDN specification itself covers privacy and security considerations on authenticated inboxes and personally identifiable information. As for persistence and retrievability, any consumer application can potentially benefit from being able to reuse notifications if the owners of a receiver makes the necessary commitments. There are further considerations which may be of interest to implementers, or may constitute future work around extending the core protocol, which are worth outlining here.

Subscribing to Notifications: The interaction between consumers and receivers describes a pull mechanism. A subscribing mechanism in which consumers request that receivers push content changes to them is left out of scope. Much of the related work requires notifications to be explicitly solicited to trigger sending. Since in a decentralised model, receivers may not be aware of possible sources for notifications, our sender-receiver relationship depends on the sender’s autonomy to make such decisions by itself. This does not preclude the scenario in which a receiver may wish to solicit notifications from a particular sender, but as there are already subscription mechanisms in wide use on the Web eg. ActivityPub, WebSub, The WebSocket Protocol (RFC 6455), Generic Event Delivery Using HTTP Push (RFC 8030), and can be interoperably combined with LDN, we do not need to specify it as part of our protocol. A push based interaction can still be arranged between LDN applications, where it would be implementation specific. For example, in a notification, a sender specifies the location of an inbox where it can receive updates about content changes. Then, an application that consumes the notifications manages its own subscriber list, and takes the role of sender in order to push content to the inbox as specified in the notification. Another specification-level integration of LDN pertaining to subscribing is that ActivityPub’s Follow Activity is a notification requesting to be notified of target’s activities when they are created. Once the request is approved, the requesting actor is added to target actor’s Followers Collection – a list of actors (with their own inboxes). The server then dispatches notifications to the inboxes of these actors.

Semantic Organisation of Notifications: While LDN facilitates a decentralised architecture for notification exchange, it does not prescribe the structure and semantics of the notifications themselves. The composition of the notifications are unspecified to foster different kinds of applications to be built to communicate information in any domain. Consequently, organisation of notification models materialises through the fact that domain-centric applications create and consume what is meaningful to them in order to coordinate among themselves. So while the notification system is globally decentralised, it enables communication that can be specialised and be only useful and meaningful to applications that understands a particular message’s semantics.

Inbox Paging: We need to consider both the needs of software systems and humans when large amounts of notification data are being generated and shared between diverse applications which may be operating without knowledge of each other. To organise and manage large amount of notifications over time, mechanisms should be in place to break representations of collections of notifications into multiple paged responses that may be easier to consume by applications.

Updating and Deleting Notifications: Receivers may want to carry out resource management or garbage collection, or permit consumers or other applications to do so. For example, an application to consume messages might let an authenticated and authorised user ‘mark as read’ by adding a triple to the inbox or notification contents.

Social Implications: In an ecosystem where anyone can (technically) say anything about anything, we can acknowledge the social challenges surrounding that in context of notifications. As the LDN protocol facilitates distribution and discovery of information, it can amplify misinformation and (community-centric) inappropriate content, as well as online and offline social dangers effectively. That is, consumers of notifications can be exposed to such information if an inbox is not managed properly. LDN may offer one way to mitigate this problem in that, the actors that control an inbox can decide on which notifications can persist and which to be removed, thereby one way to eliminate references to undesirable content created by other parties. At the same time, the same privilege can also be used to filter information and have bias towards certain parties. This is a form of decentralised curation in that while no single inbox can oversee the discovery of a resource (as there are alternative ways), it enables the owners of receivers to exert their authority on the notifications that makes it into their system. It is a middle-ground in that, on one hand, individuals can voice themselves by registering and making their resources accessible, owners of inboxes can choose not to provide pointers to it. I will further revisit some of these concerns in Addressing Social Implications.

I summarise the key protocols and data models that dokieli uses in this section. Later, I will describe important use cases in more detail.

dokieli implements the following client operations to communicate through LDP, WAP, AP.

Before dokieli makes a write request to a server, it triggers content-negotiation through HTTP OPTIONS and checks for Allow and Accept-Post HTTP headers in order to determine server’s preferences. In the absence of successful negotiation, dokieli proceeds to make the request as per the required parameters of LDN, WAP and AP. For example, HTTP POST with application/ld+json content type would be used for annotations and notifications, if a server’s preference is not determined.

Articles are by default use HTML with embedded RDFa. Where a server implements HTTP GET with text/html, and allows HTTP PUT for writing, the assumption is that it can allow payloads with text/html content type. HTML+RDFa is considered to be an important default for articles for being human- and machine-readable. Hence, if a server is capable of storing and serving resources in HTML, it will receive articles, annotations, and notifications in that form. Otherwise, dokieli will re-serialize the content into an alternative RDF syntax that the server prefers or is capable of handling per content-negotiation.

Due to Mixed Content implementations in Web browsers, ie. fetching of content over unencrypted or unauthenticated connections in the context of an encrypted and authenticated document, is subject to being blocked by the Web browser. Hence, a document available from https: will not be able to use the contents of a document available from http:. As a workaround, dokieli implements the notion of using a preferred proxy that a user-agent can use in order to access the contents of an http: resource. A document on https: fetching an https: resource will not use the proxy.

An actor may require to be authenticated and authorized with their WebID to create Web resources. The table on unit registration and content in dokieli includes the functions in dokieli that requests write-operations to a server for articles, annotations, and notifications. These client-triggered mechanisms and methods applied to creating resources are done with HTTP URIs, and helps fulfill the registration function of scholarly communication.

These client-triggered mechanisms and methods applied to creating resources helps fulfill the registration function. All created resources are considered to be archive-friendly ie. through the application of Separation of Concerns and keeping Human- and Machine-Readable Information where possible.

Authentication: Actors can authenticate themselves using WebID-TLS and WebID-OIDC mechanisms and then the server determines their level of authorization to read, write, and control resources. The workflow for authentication starts by a user inputting their WebID (HTTP URI) or their Identity Provider URL as a shorthand for their WebID into the dokieli UI. The Web resources with which the user wishes to interact may be under any domain, whether self-hosted elsewhere, so long as they are hosted by a server application conforming to the same standards.

Profile Description: In order to adopt dokieli’s user-interface and the kinds of interactions that a user can perform, the user’s WebID (HTTP URI) is dereferenced to a WebID Profile document for the following kinds of information:

• Name (eg. foaf:name and equivalents)
• Image (eg. foaf:image and equivalents)
• Homepage (eg. foaf:homepage and equivalents)
• Social graph (foaf:knows)
• Storage (pim:storage)
• Inbox (ldp:inbox, and alias as:inbox)
• Outbox (as:outbox)
• Preferences (pim:preferencesFile, solid:publicTypeIndex, solid:privateTypeIndex)
• Preferred proxy (solid:preferredProxy)
• Multiple identities (owl:sameAs)
• Additional information (rdfs:seeAlso)
• Delegated agent (acl:delegates)

When dokieli encounters owl:sameAs and rdfs:seeAlso relations in WebID Profiles, it traverses the unique connections continuously to mine for relevant information. The reason for this is both to discover a graph that can be used to adapt the interface and the possible interactions. For example, driven by follow your nose type of exploration, a human-readable name, an inbox relation, list of contacts, and other information can be put together by looking at the interconnected identity graph that is semantically associated.

Annotation of artifacts facilitates a social behaviour in that it is a collaborative effort: some actors create primary resources while others create associations between those resources, add new information, or extend existing information. Consequently, such activities help actors to learn, discover or explore new knowledge. Annotations of analogue as well as digital resources have been historically essential to sociality, education, stronger engagement, as well as being a valid form of user-generated content as per background provided in Social Annotations in Digital Library Collections, Gazan, 2008.

dokieli implements the following use cases from the W3C Interest Group Note, Digital Publishing Annotation Use Cases:

• 2.1.1 Comment on Publication Title
• 2.1.2 Tagging a Publication
• 2.1.3 Structured Review of a Publication
• 2.1.6 Annotated Resource is an Annotation
• 2.1.7 Annotation Metadata
• 2.1.8 Annotation has Multiple, Independent Comments and/or Tags
• 2.2.1 Bookmarking Current Reading Position
• 2.2.2 Highlighting a Span of Text
• 2.2.3 Commenting on a Span of Text
• 2.3.2 Cross Version Annotations
• 2.4.2 Persistence of Annotations
• 2.4.4 Annotation (or Part) is not Published Openly
• 2.4.5 Publication (or Part) is not Published Openly

A Web Annotation can be created to associate a node or a selection with a particular user motivation. The annotation can be registered and stored as a unit of information at actor’s storage, type index location, outbox, or article’s annotation service. The users can provide a text note for their annotations, assign a license, as well as designate an inbox for their annotation, so that it can receive its own notifications. The Table on WA motivations includes the method applied, location of the annotation, and associated notification.

An brief example of a Web Annotation with the reply motivation:

<http://example.org/annotation>
  a oa:Annotation ;
  oa:motivatedBy oa:replying ;
  oa:hasTarget <http://example.net/article> ;
  oa:hasBody <http://example.info/note> .

The Figure on private annotations serves to explain how each CERN requirement is realised with a standards-compliant Web application:

<article about="#a112cb4b-6814-4687-933d-8c25c917c746" id="a112cb4b-6814-4687-933d-8c25c917c746" typeof="oa:Annotation" prefix="rdf: http://www.w3.org/1999/02/22-rdf-syntax-ns# schema: http://schema.org/ dcterms: http://purl.org/dc/terms/ oa: http://www.w3.org/ns/oa# as: https://www.w3.org/ns/activitystreams# ldp: http://www.w3.org/ns/ldp#">
  <h3 property="schema:name">Sarven Capadisli <span rel="oa:motivatedBy" resource="oa:replying">replies</span></h3>
  <dl class="author-name"><dt>Authors</dt><dd><span rel="schema:creator"><span about="https://csarven.ca/#i" typeof="schema:Person"><img alt="" height="48" rel="schema:image" src="https://csarven.ca/media/images/sarven-capadisli.jpg" width="48" /> <a href="https://csarven.ca/#i"><span about="https://csarven.ca/#i" property="schema:name">Sarven Capadisli</span></a></span></span></dd></dl>
  <dl class="published"><dt>Published</dt><dd><a href="#a112cb4b-6814-4687-933d-8c25c917c746"><time datetime="2019-02-18T00:30:11.101Z" datatype="xsd:dateTime" property="schema:datePublished" content="2019-02-18T00:30:11.101Z">2019-02-18 00:30:11</time></a></dd></dl>
  <dl class="rights"><dt>Rights</dt><dd><a href="https://creativecommons.org/licenses/by/4.0/" rel="dcterms:rights" title="Creative Commons Attribution 4.0 International">CC BY 4.0</a></dd></dl>
  <dl class="canonical"><dt>Canonical</dt><dd rel="oa:canonical" resource="urn:uuid:d82804f8-e0ef-4a1e-95cd-4b2c8383eaf0">urn:uuid:d82804f8-e0ef-4a1e-95cd-4b2c8383eaf0</dd></dl>
  <dl class="target"><dt><a href="https://www.w3.org/History/1989/proposal.html" rel="oa:hasTarget">In reply to</a> (<a about="https://www.w3.org/History/1989/proposal.html" href="https://www.w3.org/History/1989/proposal.html" rel="oa:hasSource" typeof="oa:SpecificResource">part of</a>)</dt><dd><blockquote about="https://www.w3.org/History/1989/proposal.html" cite="https://www.w3.org/History/1989/proposal.html"><div rel="oa:hasSelector" resource="#a112cb4b-6814-4687-933d-8c25c917c746-fragment-selector" typeof="oa:FragmentSelector"><dl class="conformsto"><dt>Fragment selector conforms to</dt><dd><a content="" lang="" property="rdf:value" rel="dcterms:conformsTo" href="https://tools.ietf.org/html/rfc3987" xml:lang="">RFC 3987</a></dd></dl><dl rel="oa:refinedBy" resource="#a112cb4b-6814-4687-933d-8c25c917c746-text-quote-selector" typeof="oa:TextQuoteSelector"><dt>Refined by</dt><dd><span lang="en" property="oa:prefix" xml:lang="en"></span><mark lang="en" property="oa:exact" xml:lang="en">One must be able to add one’s own private links to and from public
information. One must also be able to annotate links, as well as nodes,
privately.</mark><span lang="en" property="oa:suffix" xml:lang="en"></span></dd></dl></div></blockquote></dd></dl><dl class="renderedvia"><dt>Rendered via</dt><dd><a about="https://www.w3.org/History/1989/proposal.html" href="https://dokie.li/" rel="oa:renderedVia">dokieli</a></dd></dl>
  <section id="note-20338295" rel="oa:hasBody" resource="#a112cb4b-6814-4687-933d-8c25c917c746-note-20338295"><h4 property="schema:name">Note</h4><dl class="language"><dt>Language</dt><dd><span content="en" lang="" property="dcterms:language" xml:lang="">English</span></dd></dl><dl class="rights"><dt>Rights</dt><dd><a href="https://creativecommons.org/licenses/by/4.0/" rel="dcterms:rights" title="Creative Commons Attribution 4.0 International">CC BY 4.0</a></dd></dl><div datatype="rdf:HTML" lang="en" property="rdf:value schema:description" resource="#a112cb4b-6814-4687-933d-8c25c917c746-note-20338295" typeof="oa:TextualBody" xml:lang="en">This annotation is independently created by an actor with a WebID and stored at their preferred location.</div></section>
</article>

Once an annotation is created and has a dereferenceable HTTP URI (subject to access control rules), a notification can be sent to the inbox of the target resource about the annotation if any part of an article advertises an LDN inbox. Depending on the motivation of the annotation, the notification is modelled with the Activity vocabulary eg. the Create activity (as:Create), to indicate that the actor has created the object.

In order to discover and view previously created annotations associated with an article, dokieli follows LDN’s discovery mechanism for notifications about annotations, and then subsequently fetches the annotations in order to display them in context of their target selector.

In 2016, the Annotating All Knowledge Coalition derived user stories based on the notion of what open annotation can enable. Below is a description of how dokieli realises the user stories derived from A Coalition for Scholarly Annotation. The term publisher that is used in these user stories usually refer to an entity that is other than the creators of the content. Whereas in dokieli, a publisher is any actor capable of triggering the registration of resources and having their content potentially accessible to anyone.

As an author of a paper, I want to invite a small group of colleagues to provide feedback.

Implemented with the Share feature using the actor’s social graph data, and LDN for dissemination.

As a pre-print publisher, I want to create and manage an overlay journal.

It is possible to open and render a collection of activities eg. published articles, which can be treated as a journal volume.

As a journal publisher, I want to implement a peer review system that brings authors and reviewers together in role-defined groups.

Actors can notify each other and annotate resources. However, there is no particular system to manage it as a whole.

As a university librarian, I want students’ dissertations to be annotated openly so they carry forward an expectation that things should work that way.

Annotations with various motivations are implemented.

As a biomedical researcher, I want to use human curation of the literature to develop examples we can use to train our text mining algorithms.

Actors can select, add, and edit semantic markup (HTML+RDFa) as well embed structured data blocks for Nanotations using TriG, as well as JSON-LD and Turtle. Common document formatting features are supported.

As a journal publisher, I want to create safe spaces for commentary so that authors and expert readers can participate comfortably.

Actors with the required access control privileges to an inbox manages the information flow. There is no overarching system to create or manage a space as each resource is individually controlled by their owners. It is possible for LDN inbox owners to filter notifications which refer to unwanted annotations.

As a vision-impaired user, I want annotation tools to be accessible to me.

WCAG are applied, and ARIA is used where applicable. The application of ATAG is in progress.

As an annotator, I want to be able to assign DOIs to individual contributions, or to sets of them.

DOI generation or DOI registration is out of scope. Moreover, current DOI registration systems do not have authentication mechanisms that allows annotators to identify themselves with their WebID in order to assign a DOI.

As an author, I want annotation to integrate with my preferred writing tool so that annotations can flow back into the text.

Annotations that are generated use the Web Annotation Data Model. Web Annotation Selectors and States is used to create selectors that an actor can use.

As a scientist, I want to participate in a journal club that travels around the web, annotating papers in many different publishing systems.

Any accessible URL rendered as HTML (including different publishing systems) can be annotated, and the annotation can be stored at actor’s preferred location eg. community’s annotation service.

As an author, I want to annotate my own published work, noting errata and supplying new context.

Authors can annotate with various motivations and include codetags.

As an editor, I want to document key editorial decisions and expose them to the public.

Editors can annotate with various motivations.

As a scientist I want to annotate everywhere using my ORCID identity.

Scientists using their ORCID URI (as their WebID) including the storage location in their profile can annotate anywhere.

As a researcher, I want control over the annotations I make in various contexts so I can assemble this work into a coherent form and claim credit for it.

Researchers with access controls to their annotation storage can.

As a publisher, I want to commission expert annotators whose commentary adds value to my content.

Publishers of articles and annotations can assign different inboxes to receive notifications about commentaries.

As a user, I want to be able to upvote or downvote annotations so I can help improve the quality of discussion.

Annotations can be created with the assessing motivation, as well as indicate their approval, disapproval, questioning with comments.

As a funder, I want to bring grant reviewers together in private annotation spaces where discussion is tightly coupled to the documents they are reviewing.

Reviewers can annotate with various motivations, including approval, disapproval, questioning with comments.

As publisher, I want to be able to review and moderate annotations in my own authoritative layer.

Actors can create their reviews where they prefer and have access to. Moderating annotations owned by other actors can only be done if the publisher has access to.

As an author, I want to be able to annotate and interconnect the references I cite.

Citing and annotations with various motivations is implemented.

As a publisher, I understand that readers may view my content through a variety of lenses, but I want to present my own annotation layer as the authoritative one.

Annotations can be created with the commenting motivation, which gets embedded into the document (achieved by those with write access).

As the developer of a journal publishing system, I want to deliver a bundled annotation service that integrates with my identity system.

Using open standards for profiles, authentication and annotations means that anyone wishing to provide a hosted service need only adopt the same standards, eg. WebID, OIDC and WA.

As a publisher, I want to project a curated view of a paper’s peer review process onto the published paper.

Memento and provenance information is generated and consumed in articles. All annotations and unique inboxes can be used. Robust Links are also shown.

A screencast of an annotation in dokieli is shown below.

The mechanism for personal identities, social graphs, and self-publishing articles and annotations enables a new layer of decentralised interactions. Here I describe how actors can discover units of information created or annotated in response to Web resources (like articles) by their social network without intermediaries (which includes the authors of the article or the features supported by the server it is published at). This scenario expands on identity A knows identity B, and identity B has a storage or outbox as declared in Linking the Decentralised Information Space.

As each unit of information can be independently published and made accessible by an actor, one way to discover the information is by determining the relation between their WebID and the location of their collection of artifacts eg. under their personal storage (pim:storage) or activity outbox (as:outbox). The second connection between the actor that is interested in discovering such information and the actors which created them (foaf:knows). Hence, we are able to discover the reactions (like replies, comments, approvals, questioning, or annotations in general) made by the members of an actor’s social graph.

In order to realise this scenario, the required components are as follows:

• Actor A wants to read annotations or activities around a unit of information.
• Actor A knows actor B.
• Actor B has a collection of artifacts.
• Actor A has read access to Actor B’s activity collection.
• Actor A’s application associates the unit of interest with actor B’s annotation or activity.
• Actor A’s application displays the annotation in context of the unit of information.

As these components are solely based on interoperable standards, any conforming application is able to interact with them, and perhaps more importantly, is not subject to vendor lock-in.

This scenario is realised in dokieli, and works as a single-page application or via the Web browser extension. The Web Extension enables a new dimension to discovering and consuming independently published activities in that, a user can navigate to any HTTP URL and potentially discover and interact with annotations made by the members of their network. Thereby, a layer of decentralised and linked research space can exist independently, where any Web resource, perhaps typically an article, can be incorporated into the social or research discourse.

As each annotation (or unit of information) can have its own inbox, subsequent interactions with the annotation can be both stored at the annotator’s storage and notified at the advertised inbox. This further decouples the components required to publish and consume content on the Web in an interoperable way. The decentralised activities in dokieli demonstrates the utility of independent publishing, semantic relations, as well as extensibility of structured Web objects.

A key aspect of the Social Web is sharing our creations and interests with the others. After (optionally) authenticating with a WebID, dokieli documents can be shared with contacts, which are discovered from the user’s WebID profile. Contacts whose profiles advertise an LDN Inbox will receive a notification of the share. The notification is modelled with the Activity vocabulary, eg. Announce, to indicate that the actor is calling the target’s attention to the object. Recipients can use any LDN-compatible application to view the notification, without needing to have ever used dokieli before.

Publishing and Referencing and Scientific Expressions described some of the existing specifications to identify, integrate, and link assertions in research. The registration of units of information with URIs enables us to create RDF relations between citing and cited entities. For example, actors can link to specific statements within and across resources, as well as annotate their intentions while preserving the context when they are created.

Contextual Citations: One way to create a citation in dokieli is by authors selecting a node or a text fragment of interest in their document, and then inserting the URI of the resource to be linked to, as well as a specifying the type of semantic link between them in a specific or general way using the classifications available from the CiTO ontology eg. using agrees with, confirms, cites as evidence. The input for the cited entity can be any HTTP URI, DOI string, or ISSN string, and when resolved from a corresponding HTTP URL, the content in RDF is used towards constructing the citation. dokieli takes the input target URI, and retrieves data (like title, authors) expressed in RDF in the linked document, integrates a statement expressed in RDF eg. figure cites as evidence, in context of the content, the citation intention as human-readable text, and adds a scientific endnote reference including the retrieved information. As a citation results in the assigning an HTTP URI for the citing entity, it can be further semantically annotated by the actor.

<http://example.org/article#argument>
  cito:citesAsEvidence <http://example.net/observation#results> .

Robust Citations: In order to preserve the context of the citation as well as taking content-drift and link-rot into account, first a snapshot of the cited entity’s URI is created at the Internet Archive, and then both the archived URI and its datetime stamp is incorporated to the referencing hyperlink – the original citation resource.

Notifying Cited Entity: Once a citation is made through the dokieli user interface, dokieli proceeds to check if the cited resource advertises its own LDN inbox, and if so, sends a notification indicating the citation function. This notification serves as a back-reference announcement, ie. being informed that a citation was made and thus creating discoverable bi-directional linking between the units of information. With the help of an inbox consumer, like dokieli, the cited article can also inform the reader about the back-references out there on the Web.

Registrations: The citation activity results in the registering HTTP URIs for the citing resource, archived snapshot, and the notification resource.

The following screenscast illustrates a clientside interaction in dokieli where the user cites a unit of information with a particular relationship, and a notification is sent to the cited entity’s inbox.

It is possible to use dokieli’s create operations as a composite to fulfill social and scholarly activities. One common scenario includes soliciting reviews, reviewing, announcement of the review, and final decisions made by community members. Core mechanisms, annotations and sharing resources in dokieli is as follows:

• Actor A publishes their article at preferred location.
• Actor A sends Actor B a notification requesting a review of their article.
• Actor B creates a derived copy of Actor A’s article at their preferred location. The derived resource optionally includes its own inbox and annotation service.
• Actor B reviews and annotates the derived resource by storing their annotations at preferred location. Notifications are sent to the inbox about the annotation.
• Actor B archives their derived copy with annotations.
• Actor B sends Actor A a notification sharing the location of the annotated resource.
• Actor A visits Actor B’s derived resource and discovers the review and annotations.
• Actor B sends a notification to a community inbox that is dedicated to monitor scholarly activities about their review and annotations so that it can be discovered by other actors.
• Actor C discovers Actor B’s review through the community inbox.

This particular scenario shows that it is possible for each actor to create and control resources, including articles, annotations, and notifications at their PDS or preferred locations. It is one way to combine standards-compliant mechanisms where actors can shape their content, self-publish, as well as share their works with other actors and services. In essence, the registration, certification, awareness, and archiving functions in scholarly communication are fulfilled in this scenario without a central authority.

Creating awareness and enabling actors to protect their data privacy is essential to fostering user’s trust in the systems they use. The adoption of standards and practices with different degree of control can yield different privacy measures. In the decentralised and interoperable environment that dokieli operates in, there are a number of challenges and approaches. I discuss user privacy, data privacy, preferred proxy, and private activities below. The terms anonymity, unlinkability, undetectability, unobservability, and pseudonymity follow the definitions from A terminology for talking about privacy by data minimization: Anonymity, Unlinkability, Undetectability, Unobservability, Pseudonymity, and Identity Management, Pfitzmann, 2010.

User Privacy: dokieli acts as a conduit to perform read-write operations against PDS. As the rules for read-write operations are set by servers, dokieli does not require the creation of user accounts at any server, in order to interact with the server’s resources. Where applicable, actors can participate by maintaining their anonymity, pseudonymity, or can choose to identify themselves to dokieli by using their preferred personal identities, without necessarily disclosing "real" information. Actors authenticate against the servers they prefer in order to perform access-controlled actions. It is possible for the environment in which dokieli operates in to transmit device, machine, or browser fingerprint, thereby potentially risking unlinkability – however, they are outside the control and scope of dokieli.

Data Privacy: With respect to data privacy, actors can exercise their autonomy through the use of PDS for their profiles, articles, annotations, and notifications. As a PDS can provide us with the means to safeguard data, dokieli as a clientside application enables actors to perform read-write Linked Data operations where they prefer under their own conditions. With the help of access controls, a PDS that dokieli communicates with can potentially fulfill all minimizations – anonymity, unlinkability, undetectability, unobservability, pseudonymity – possibly with the exception of the domain name and IP that the PDS operates from.

Preferred Proxy: A proxy acts as an intermediary between the clientside application and the destination server. This is useful in a number of scenarios for dokieli, such as getting around Mixed Content issues, absence of Cross-Origin Resource Sharing (CORS) support, as well as concealing the identity of the actor, fulfilling anonymity, unlinkability, pseudonymity, when making requests to a target server. An actor’s preferred proxy can be determined by checking their WebID Profile description for the solid:preferredProxy relation, and if found, dokieli can use it to route requests when needed. Alternatively, a built-in (but user configurable) proxy location in dokieli is used.

Private Activities: The ability for actors to create and discover activities made within their social network without going through intermediaries, makes it possible to isolate units of information that is both access-controlled and invisible (or be unknown) to third-parties. These class of interactions in dokieli is aligned with the notion of undetectability and unobservability.

dokieli implements sparqlines: statistical observations fetched from SPARQL endpoints and displayed as inline-charts. An inline-chart, also known as a sparkline, is concise, and located where it is discussed in the text, complementing the supporting text without breaking the reader’s flow. For example, the  GDP per capita growth (annual %) [Canada] claimed by the World Bank Linked Dataspace. dokieli demonstrates how scientists or authors can better support their argumentation with sparklines generated from their own or public statistical linked datasets.

dokieli is a clientside application that uses the user-agent to interact with server applications. It has two deployment approaches including the same set of features: single-page application and Web browser extension.

dokieli’s presentational and behavioural code layers can be included in Web "pages" in order to trigger them as active single-page applications. It is a smart client that allows different kinds of articles eg. academic, blog posts, news, to be authored and annotated from within Web browsers, without necessarily having them deployed from a server, ie. it can be used offline or on localhost. dokieli internally handles its content and structural and semantic representation based on user’s actions and available information. Articles, profiles and their contact information, notifications, annotations with different motivations, for instance, can be read and written ubiquitously to any Web space with supporting HTTP methods and access control mechanisms.

The Web browser extension is a thin wrapper around dokieli’s core code in order to embed itself in any HTML-based Web page on the Web. It inherits all of the features of a single-page application. While HTML based documents on the Web vary in their quality, dokieli’s write operations generate HTML+RDFa. One of the utilities for the extension is to have a consistent interface for annotating node or text selections on a Web resource, building structured citations, creating derivations of Web resources as well as sharing document fragments with ones contacts via notifications, without having a service dependency or being limited by the Web page’s UI.

Dimensions

Subject - the test subject of an assertion

Test - the test criterion of an assertion

Mode - the mode in which the test was performed

Measures

Result - the result of an assertion (Outcome, Info)

Identifier

a2ebd5be-b9ca-11e8-bdaa-338eda30385e

Published

2018-09-16

Creator

https://csarven.ca/#i

Structure

FAIR Metrics