Would you do me a favour? I'd like to stop talking for a minute and when I do, take a look at the room you're in and above all at the man-made objects in that room that surround you – the television set, the lights, the phone and so on – and ask yourself what those objects do to your life just because they're there. Go ahead.

Well, that is what this thesis is going to be all about. It is about modern research communication, and just because it is there, shapes the way we think and behave; and why it exists in the form it does, and who or what was responsible for it to exist at all. What is an alternative way of change?

The idiosyncratic relationships between technological breakthroughs and societal transformations throughout history are intrinsically intertwined. It was not until I came across Connections, the TV series by James Burke that I had a glimpse of the phenomenon about the increasingly interlinked human endeavours over time. After encountering Marshall McLuhan’s theories on communication and how media have the power to shape and transform human nature, what to do next was mostly clear (in my head). Given these realisations or perspectives as a foundation, being a Web technologist only enabled me to build the necessary connections between what was previously missing or underdeveloped. The rest was mostly a matter of struggling with the ghost in the machine and working alongside our shared social challenges.

Ironically, even today, academics essentially operate within a paper-centric framework to create and disseminate publicly-funded knowledge by usually delegating it to third-party publishers which still operate in the 15th century. Meanwhile, the Web – if we can anthropomorphise for a moment – is disappointed by the distracted academics’ practices. I took the liberty to test the boundaries and demonstrate some of the aspects of what the native affordances of the Web provided – an extension of our central nervous system.

I live in an information society, which makes use of variety of communication media to meet societal needs. The Web has shaped social policies and practices around the world, as well as influence the creation and use of other technologies. It has also reconfigured individuals to be active participants in global information exchange, as well as passive consumers partly due to the abundance of instantaneously available multimedia. To date – 30 years old as of this writing – the Web is considered to be a net positive for society all meanwhile open and complex challenges remain. Like any human technology, what we devise and how we use them has societal implications.

We can start this story from anywhere. I argue that in order to best contextualise this thesis, it is essential to operate under the understanding of communication mediums and scientific revolutions, as well as their effects on society. This is especially because the research goals of this thesis at its core is sociotechnical, and that a clear division between them will not do its justice. I believe that in order to push the boundaries of our knowledge further, it necessitates a multidisciplinary undertaking – taken with a grain of salt. One overarching goal is to foster the social machinery from different perspectives that is required to build a knowledge Web for any type of user; human, machine, or other. While I set the problem context to scholarly communication, applications of the knowledge and generated artifacts are potentially applicable to other initiatives in society.

I assume that researchers in scholarly communication are motivated to some extent personally, whether that is with the goal to take part in advancing collective knowledge, society and life, career advancement, prestige, or financial gain. Researchers are intrinsically interested in making their impact by way of sharing their findings in a manner that is most accessible to potential users. As for the underlying technical machinery for information exchange, I assume that it is useful to aim in designing interoperable information systems to help us address problems in context of societal goals. All of these aspects are part of the social machine of scholarly communication, and are explored through the sections.

As many have argued, for scholarly communication to be more effective; accessible, usable, inclusive, as well as equitable, we need to adopt methods that are better at identifying and making use of the native affordances of the Web. This is particularly important today since academic and social activities use the Web as the primary medium to communicate (in contrast to systems that mimic paper media). Hence, I examine the core characteristics of the Web, discuss the state of affairs in scholarly communication, and share my findings on how a way to fulfill the forces and functions in scientific communication can be met by a socially-aware decentralised client application along with an ocean of open Web standards at its disposal.

This section contains the outlines for social and technical research goals of this thesis. Each goal is further explored and related material is reviewed in relevant sections.

I first outline the Technical Research Problems and then a set of corresponding Knowledge Questions. The questions refine the Research Goals.

Static stand-alone paper is not an appropriate habitat for a thesis on linking. Therefore, the paper is included with the CD-ROM, not vice versa.

All information of significance in this thesis is made available as Linked Data on the Web and archived.

The effects of technological mediums on society have some similarities to the evolution of scientific paradigms. This will serve as the historical framing for this thesis and helps to contextualise the application of core methods and values of the Web on scholarly communication.

Next, to continue to paint the historical backdrop against which this thesis is placed, I look at the European-centric scientific revolution which overlapped with the transition from the manuscript to the print-centric mode of information organisation and dissemination.

We need to examine our assumptions about scholarly communication on the Web in order to understand its social implications. Building on the previous brief histories of media evolution and scientific communication, the next review is of the history of the Web (a social machine) and then specific role the Web plays in contemporary scholarly communication. The focus of the rest of this section is on structural changes in scholarly communication, while the evolution of technical advancements runs in parallel.

Choice is an illusion, created between those with power, and those without.

In Mediums and Paradigms I have discussed the wide reaching effects of communication mediums as well as scientific frameworks adopted by society. In this section I zoom in on the history of scholarship in order to contextualise the practices and their influences to date.

As the print age is an important historical backdrop of this thesis, I provide a brief review of The Printing Press as an Agent of Change to further contextualise social transformations and the systematisation of scientific exchange given the availability of a technology for the masses:

The publication and dissemination of information – in the most general sense of the words – played a key role in the advancement of societies’ knowledge for centuries. The first academic journal, Journal des sçavans and shortly after the Philosophical Transactions of the Royal Society in 1665, evolved out of the exchange of letters among scholars and the transcripts of the meetings in scholarly societies discussing experiments, publications and natural curiosities. With this in mind, the most widely accepted purpose of a scholarly article is a way to communicate research results and other intellectual work with the public; to have an accurate archival record of the research findings; to build on top of earlier knowledge; as well as to have it serve towards crediting authors for their work. Today’s scholarly journals, containing quality-controlled research reports, closely resemble the first journal. Encyclopedia of International Media and Communications, Scholarly Journals, Vol. 2, p. 688, 2003, states that due to the growth and demand for copies of the literature from non-members, mostly institutions, the larger journals began to sell subscriptions at relatively higher prices than members pay.

In The Sociology of Science: Theoretical and Empirical Investigations, Merton, 1973, discusses four sets of institutional imperatives taken to comprise the ethos of modern science, originally introduced in The Normative Structure of Science, Merton, 1942:

• Communalism: in order to promote collective collaboration, intellectual output should have common ownership.
• Universalism: scientific validity is kept independent of political and personal factors such that everyone has equal rights and possibility to obtain information.
• Disinterestedness: the practice of science should benefit the common scientific enterprise with no conflicts of individual interest or gain.
• Organised skepticism: scientific claims should be exposed to critical examination before being accepted to be part of common knowledge.

I now discuss several movements which emerged since the 1990s towards fundamentally transforming scholarly communication using newly available digital network technologies.

The scientific communication market is an interplay between actors with an objective to exchange units of information along the lines of: academic literature, analysis, annotations (eg. assessments, bookmarks, comments), datasets, evidence, interactive representations, software, observations, provenance, and workflows. Further, researchers commonly want to expose their work beyond the scientific community to the wider world, seeking adoption or contradiction of new ideas and findings.

In the seminal work, Forces and functions in scientific communication, Roosendaal, 1997 provide a framework which has had historical influence on research and development pertaining to technical challenges on interoperability, workflows, service sharing, and information modeling in scholarly communication. Through this thesis, I will refer back to these concepts as a means to describe and compare various technologies and approaches to scientific publishing.

Roosendaal and Geurts disassemble the scientific communication market as an interplay between four forces and functions. The dynamics between the forces and functions is described as a tetrad:

Forces

The scientific communication market consists of four forces with each having complementary parts: actors, accessibility, content, and applicability.

Actors

Any kind of interactive agent that creates or uses information, eg. authors, readers, tool implementers, policy makers, application programs.

Accessibility

The availability and retrievability of information, content accessibility.

Content

The generated questions and answers by actors.

Applicability

Transfer of knowledge to science, technology, and society.

Functions

Scientific communication consists of four functions: registration, awareness, certification, and archiving (RACA).

Registration

Allows claims of precedence for a scientific activity.

Awareness

Remaining informed of the observations.

Certification

Establishes the validity of registered claims.

Archiving

Preservation of scholarly records over time.

The actor and content forces are generic and internal to the scholarly market, and are considered to be indispensable in that there would not be a scientific communication market if there are no researchers and the artifacts they exchange. The accessibility and applicability forces on the other hand are external forces because they shape, give purpose to, or influence the activities and content.

The registration and archiving are objective functions in that they are external to the research process. Information exchange cannot take place if there are no usable identifiers for the units of communication or if no knowledge preservation measures are put in place. Certification and awareness are subjective communication functions – internal to the research process. Registration and awareness are different aspects of making information findable, or scientific observations. Certification and archiving are aspects of scientific judgement, ie. they act as filtering mechanisms in the information space.

In today’s scholarly publishing, the journal article is considered to have four chief functions:

• The dissemination of information
• The archiving, or permanent preservation, of the information
• The assignment of priority in the information
• Quality control

In Rethinking Scholarly Communication, Van de Sompel, 2004, proposes to revise the notion of a unit of scholarly communication in two ways. In a technological sense, it suggests to recognise the compound and dynamic nature of a unit of communication that work on the Web. Content can be the result of a combination of other registered items; data may be available through interactions, or depend on acquiring externally registered items. In order to empower scholarly communities from a systems perspective, the article posits that there should be more flexibility in what constitutes a unit of scholarly communication in order to allow registration, as well as factoring timing and the quality of what is to be registered. The article also acknowledge the presence of an additional function in the market which reflects the academic credit and policy in the current system. A derived function, rewarding of actors for their performance in the communication system based on metrics derived from that system. For rewarding to take place, the information space needs to provide the ability to extract meaningful metrics.

In A Perspective on Archiving the Scholarly Web, Van de Sompel, 2014 characterises the ongoing evolution and transition of objects pertaining to scholarly communication. In particular, the communication is becoming more Web-centric ie. visible, continuous, informal, instant, content-driven, and the communicated objects vary in their type, have compound composition, diverse, networked, and open.

Given that the emergence of the Web influenced the communication market by transforming it from predominantly being medium-driven into a content-driven ecosystem, as well as the research process becoming more visible on the Web (per the Open Science movement), the article also posit that the aforementioned indicators impact the functions in scientific communication, and in particular have a significant impact on the future of archiving. The observation is based on the archival process held on the Web; unplanned and unconstrained archiving, organically occurring, by way of Web objects being deposited across distributed and dedicated archival services. They describe the effect of Web-based archiving with respect to the four functions:

• registration: a wide variety and versions of objects that are compound, dynamic, interrelated, and interdependent are available from different Web platforms;
• certification: a variety of mechanisms that are decoupled from registration;
• awareness: a variety of objects being communicated in real-time and social;
• archiving: no constraints on the number or kinds of parties that can hold archived copies, the need for appropriate levels of cross-archive interoperability in order to support seamless, uniform access to archived objects, that are temporally appropriate while context being preserved

The authors conclude that in order to address the needs of a distributed archival approach for scholarly Web objects, standards will play a central role for interoperable systems.

Given the dynamic nature of units of communication on the Web, vertical integration, stepping through the four functions in order (as per the traditional journal-centric ecosystem), turns out to be inadequate. Hence, an affirmation of the outlook as outlined in Rethinking Scholarly Communication, that the essential functions need to be fulfilled in discreet, disaggregated, and distributed manners, and in which a variety of networked pathways interconnect the autonomous hubs that fulfill these functions. The act of recording (or registering in the most general sense) a unit of communication is not a replacement for archiving or persistence, even if the freely available Web services are used towards this end. The authors distinguish recording and archiving in that the former is short-term, no guarantees provided, used for many read-write operations, and part of a scholarly process. Archiving serves a different function, in that it is intended for the long-term, and there is an attempt of a guarantee (or promise) and the archived artifact becomes part of the scholarly record.

The authors call for another archival paradigm that is based off the characteristics of the Web. That is, given the inherently interlinked and dynamic nature of Web objects, archived resources cannot be treated as atomic resources as they do not necessarily contain the complete content. For instance, a research document on the Web may refer to external media or interactive scripts, aggregated objects, and changing or ephemeral objects. This is in contrast to the traditional journal-centric approach to archival where all material – physical paper – can be eventually tracked to library stacks. Hence, the future of archiving needs to consider the nature of the object, including the degree to which it can be archived. The complexity and care that is required for the proper preservation of digital objects is further described in Requirements for Digital Preservation Systems, Rosenthal, 2005, provides a taxonomy of threat models: software, communication, network, economic, organisational, to name a few, which pose a particular risk to permanent information loss. Thus, it is important to bare these potential issues in mind when considering what scholarly units of information will contain and how they will be made available.

All of man’s artifacts – whether language, or laws, or ideas and hypotheses, or tools, or clothing, or computers – are extensions of the physical human body or the mind.

In recent decades there has been a global shift in communication forms used for scientific activities. This perhaps demonstrates a desire for amplification of the senses that are used to consume and create knowledge. For instance, print-centric emphasises the linearity of scientific information representation and its communication, whereas a Web-centric approach makes multimodal interactions possible within the same domain. Scientists need no longer fit their understanding of the world within the confines of static and two-dimensional descriptions of phenomena. The need to travel great distances for first-hand observations and collaboration is also reduced. With electronic media, and more specifically the Web, we have the option to also watch, listen, and participate around measured observations, instantaneously as the events occur, or after the fact around persistent and detailed logs. Just as the invention of the printing press transformed the codification and mobility of knowledge, the Web transformed and extended human and machine communication on all fronts. What will be further explored in this thesis is the extent to which our current scholarly activities make use of the inherent qualities of the Web, as well as ways to further take advantage of these qualities.

Tetrad of Media Effects: The Web is an artifact of our society. In order to help us understand the effects of the Web on scholarly communication, we can tetrad of media effects from the Laws of Media, McLuhan, 1989, as a heuristic device; a set of four questions:

• What does it enhance or intensify?
• What does it render obsolete or displace?
• What does it retrieve that was previously obsolesced?
• What does it produce or become when pressed to an extreme?

Effects of Web Centric Scholarly Communication: The questions are intended to be complimentary and studied simultaneously. They correspond to four effects: enhances, obsolesces, retrieves, and reverses into. I use this tetrad to situate the effects and artifacts of scholarly communication on the Web in relative to adjacent mediums, in particular, to print media:

Enhances

access to information, autonomy, social connections, decentralisation, distribution, immediacy, multi-modality

Obsolesces

centralisation, isolation, distance, memorisation, privacy, tactility, travel, linearity, typographical conventions

Retrieves

community, correspondence, direct representation, free expression, dialogue, involvement, public space, self-control

Reverses into

collectivism, degraded information quality, group-think, information overload, fabrication, lack of accountability, noise, third-party control

Throughout this thesis, I will refer back to these concepts.

The tetrad serves as a conceptual categorisation of impacts of a medium that a particular media affords. It allows us to outline and analyse what the medium enables or has implications for based on observations in "the real world", or in day-to-day academic practice. In particular, we can examine the Web’s effects and artifacts within with regards to composition of scholarly information, individual autonomy to create and consume knowledge, and sociality of scholarly communication. By examining these effects, we can hone in on understanding what the underlying form does to us, society at large, and scientific communication itself. It provides us with a foundation to go further in uncovering the nature of scholarly communication’s interplay with the Web.

How does scientific communication evolve through the mediums it operates from?

In the 1967 book, The Medium is the Massage, McLuhan wrote: at the high speeds of electric communication, purely visual means of apprehending the world are no longer possible; they are just too slow to be relevant or effective. While McLuhan’s critique was about the effect of the print media in the electric age, it stands especially valid today in the age of interactive new media. For instance, the World Wide Web is a cool media because its access, multi-linearity, and interactive possibilities offer multi-sensory states for individuals. This is in contrast to print, considered by McLuhan to be hot media, demanding primarily the attention of our vision. As the printed word’s requirement for the other senses is minimal, it spoon-feeds the content to the reader in a l-i-n-e-a-r fashion. The literate or print culture is passive and is detached from immediate involvement because the medium has no such requirements. The cooler media on the other hand, requires a greater level of engagement from the audience at a faster rate. This increased engagement allows us to detect change and better recognise patterns, which, according to Gestalt laws of grouping, makes the whole of the information easier to follow and understand.

The interaction of media with our senses is an inherent property of media. The media’s built-in sensory bias affects and transforms human society and psyche. Modern media, like the Web, offer extensible interactions, creative participation, and social engagement. There is a natural human desire to connect and exchange ideas, behaviour, or style with others, as well as to co-create. Given these opportunities, the fundamental question to ask here is, why should research communication be limited to, evaluated by, and rewarded on methods and practices be based on print-centric media? Scholarly communication still uses methods which date back to the invention of the mechanical movable type printing press (c. 1450 CE), even in the advent of the globally accessible and interactive new media that are at our fingertips today, and at much faster turnaround times from publication to consumption than ever before. The practices correspond to publication workflows that are in place for academic communication, and have remained largely unchanged for over a century. What does it mean to have an academic achievement through a “publication”? If we consider the Web as a potential supplement to or replacement for these older practices, and given that the Web can be purposed for greater possibilities than print alone, then we (members of the academic community) also need to address the question, what constitutes a scholarly contribution?

Today, it is evident that the typical academic publishing scenario requires adherence to the printability aspects of the information eg. page length, typographical guidelines. That is, all of the encompassing data or supporting artifacts for any given research output typically must be channelled through the characteristics of print media. In practice, this leads the research community to use tools that are tailored solely for printing or viewing (on screen). This (sticking to print) leads to unnecessary information loss or arbitrary filtering of the content, which might be otherwise preserved if a different medium was used. Since print is only one possible way to represent and communicate the information, it is worthwhile to investigate and employ alternative or complimentary representations that are more appropriate given the affordances of the Web. Traditional electronic document formats, on the hot end of the media spectrum, are trying remain relevant inside a cool medium.

This section served to acknowledge initiatives and developments that went into scholarly communication in the age of information and Web, against their historical backdrop. It also introduces some frameworks for description and analysis of technologies and processes particular to scholarly communication, and communication mediums in general, which I will continue to refer to through the rest of this thesis. Subsequent sections are dedicated to devising a scholarly space that is more fitting to the medium of the Web.

Given that global strategic research is influenced by policies and industry, different information spaces can be created by an ensemble of units of information. Such information spaces would serve to facilitate knowledge exchange in the scientific communication market. In this section, I investigate existing approaches towards configuring different aspects of Web-centric information spaces.

Scholarly knowledge includes a range of research artifacts that needs to be described, uniquely identifiable and be discoverable on the Web. These include research articles, peer reviews, research data, social interactions like review requests and notifications in general, as well as different kinds of annotations on research objects. The current state of access and use of scholarly knowledge is insufficient for society at large. By enabling accessible Scholarly Knowledge (Knowledge Graphs: New Directions for Knowledge Representation on the Semantic Web) graphs as well as applications which make use of it, we hope to enable universal access to previous research. By improving the availability through linked research, we can facilitate discovery and building on existing research. A fundamental first step is to investigate and develop effective ways to represent fine-grained information that is accessible, human- and machine-readable, and interconnected.

A typical high-level interplay between the actors and the content forces, and their functions (registration, awareness, certification, archive) in scholarly communication have authors reply to call for contributions by sharing their literature. Reviewers provide feedback on the research results; individuals annotate and re-share the literature; editors filter and assemble a collection of work; archives are notified about the research. While the document-centric Web was mostly for human-readability, Linked Data-based Web is oriented towards improving discoverability, readability, and reuse for machines, and in effect better helping humans. For example, if research articles capture their problem statements, motivation, hypothesis, arguments, workflow steps, methodology, design, results, evaluation, conclusions, future challenges, as well as all inline semantic citations (to name a few) where they are uniquely identified, related to other data, and discoverable, then specialised software can be used to verify the article’s well-formedness with respect to the domain. In essence, this materialises the possibility of articles being executable towards reproduction of computational results. Similarly, user interfaces can manifest environments where readers can rerun experiments or observe the effects of changing the input parameters of an algorithm. This has the affordance for a more involving environment for the user, improves learnability of material, and supersedes the passive mode of reading.

In a 1999 article, Practical Knowledge Representation for the Web, Van Harmelen, 1999, stated that the lack of semantic markup is a major barrier to the development of more intelligent document processing on the Web, and meta-data annotation of Web sources is essential for applying AI techniques on a large and successful scale.

In this section I examine some of the existing Web standards and practices for the structure and semantics of scholarly information with focus on narrative documents and observational data. Scholarly documents with the characteristics of a continuous prose, such as those of research articles typically includes factual information about an investigation with supporting provenance-level information and references to underlying data. Annotations in the general sense can be similar to articles, however they are generally intended to encapsulate an indirection where target resources are associated with annotation activities with motivations, eg. commenting, bookmarking, or classifying. Notifications for scholarly activities serve as a way to inform a target of interest about events in general. I also examine representation and publication of statistical, experimental, or observational data on the Web following the Linked Data design principles.

One goal here is to identify and apply patterns so that content creators can register different units of information at varying semantic granularity. Ultimately, I seek solutions that embrace interoperability and reusability, device and medium independence, as well as favourable for accessibility and archiving.

The structure of a scholarly journal article has been mostly conserved during the evolution of scientific communication. A typical scholarly article may include a sequence of sections, usually an abstract; introduction and background relevant to earlier work; methods used to justify experiments; research results; discussion and conclusions; and a list of references. Different forms of research eg. original, scientific, artistic, or research in the humanities, may have their own discipline-centric organisational structures for the journal article. For example, Introduction, Methods, Results, and Discussion (IMRAD) is a widely adopted standardised structure in original research articles as per survey, The introduction, methods, results, and discussion (IMRAD) structure: a fifty-year survey, Sollaci, 2004. In the medical domain, structured abstracts are used to help with rapid comprehension by having distinct labelled sections, eg. introduction, objects, methods, results, conclusions, within the abstract of the article.

This section covers semantic structure of the narrative aspects scholarly communication. This can enhance both human- and machine-readability of information, in particular, representations of units irrespective of how they are produced.

Non-narrative units can also be semantically represented to enable easier reuse and interpretation; this is covered in more detail in Linked Statistics and Interactive Linked Statistics.

There are plethora of approaches and developments to representing scholarly information with markup languages. Here I focus on a few and describe their characteristics.

Journal Article Tag Suite (JATS) is an XML format used to describe content and metadata of research and non-research articles eg. including reviews, editorials, instructions to authors. It can be used by publishers and archives to interchange scholarly content in a uniform way. The intent of its Tag Suite is to prescribe and to preserve the structure and semantics of the original content independent of the presentation. JATS includes three Tag Sets: journal archive and interchange, journal publishing, and article authoring. In the case of peer reviewed scholarly articles that are intended to be archived, the publication may also include information about the conference the research was presented at, data about the publication and journal, document history, license, funding, and various kinds of annotations. Given that JATS is XML-based, different manifestations of the original content are generated through transformations eg. as a print-version, HTML, as well as RDF as described in From Markup to Linked Data: Mapping NISO JATS v1.0 to RDF using the SPAR (Semantic Publishing and Referencing) Ontologies.

Text Encoding Initiative (TEI) has guidelines to define and document a markup language for representing the structural, renditional, and conceptual features of texts. TEI is similar to JATS in that it aims to improve understanding of text made explicit with XML, as well as for information interchange, integration, and preservation. Its encoding methods are mainly aimed at machine-readable texts in academic, linguistic, literary and technical documents. Its document structure is similar to JATS in that it has a front (header), body (article), and back pattern, and including a generalistic approach to coping with text with any size, complexity, writing system, language, date, or media. For instance, the specification helps with information identification (eg. page numbers), structural divisions, pictures and diagrams with captions; different writing modes (eg. prose, verse); each with formal structural units (eg. paragraphs, lists, stanzas); with textual distinctions (eg. titles, names, quotations); as well as metatextual indications (eg. corrections, annotations, revisions).

Electronic Publication (EPUB) is implemented as a distribution and interchange format based on XML and Web Standards. EPUB is delivered as an archive file (ZIP) packaging XHTML, CSS SVG, along with supporting media files. The EPUB Content Documents defines profiles – inheriting XHTML and extensions to the underlying document model – for the content to be used in the context of EPUB Publications, and EPUB Open Container Format defines the file format and processing model for the single-file that encapsulates all material. Software on devices complying with EPUB can read and interact with its content. It retains the reflowable features of XHTML and CSS in that content presentation can be adapted by a consuming device. The specification allows embedding of RDFa attributes.

Research Articles in Simplified HTML (RASH) is a format accompanied with the RASH Framework for writing HTML-based scholarly papers. RASH aims at sharing scholarly documents through the Web while working within the existing publishing workflow, and is expected to be produced from MS Word, ODT and LaTeX sources. RASH has been developed as a formal grammar using RelaxNG which includes a subset of elements – 32 – from HTML5. With the exception of the script element, it is compiled of printable elements. Its grammar facilitates RASH documents to run against a markup validator, to convert between sources, rendering for Web and print media, and to extract additional semantics. Its rationale for representing human-visible and machine-readable information in HTML varies eg. title, authors, keywords of an article are used as hidden metadata, and require additional processing to make it human-visible in the Web browser, as such a dependency on JavaScript support and being enabled. The resulting grammar favours simplicity towards authoring over expressing accurate or appropriate semantics eg. h1 for all heading levels is used, as opposed to HTML5’s required heaving levels (hX). The format allows DPUB-ARIA and RDFa annotations, as well as embedding JSON-LD, Turtle, and RDF/XML data islands with the script element.

W3C Scholarly HTML Community Group has its mission to build a common, open format for the exchange of scholarly information. Its specification on Scholarly HTML proposes a vernacular document format aimed at encoding scholarly articles built on open standards, as well as having them compatible with off-the-self Web browsers. One of its high-level goals is to enable structured metadata as well as semantic enrichments, accessibility, internationalisation. It applies the notion of semantic overlays for its markup patterns with focus on role-based semantics as defined in WAI-ARIA, DPUB-ARIA, and semantic representations using RDFa. The – work-in-progress – specification explains how concepts for people and organisation, article semantics, schema roles, actions, citations, document rights, and so forth can be marked using appropriate vocabularies, using the schema.org vocabulary as base. In order to simplify authoring, where applicable, it is encouraged to use RDFa markup patterns where the depth (level of nesting in the DOM tree) for the RDF statements would be relatively flat. One consequences of this rule of thumb may be that there can be duplicate information or parts of the information being human-visible whereas some others would be intended only for machine use.

Summary of Characteristics: For historical and technical reasons, here I summarise some of their (shared) characteristics:

• With the exception of Scholarly HTML, the markup methods first cater to printability (as opposed to being Web-centric) or require transformations to other formats in order to be viewed and interacted further.
• Data representations are prescriptive in that they are coupled with specific applications or workflows.
• The generation or use of markup is dependent on specific environments for use or interchange, media or devices.
• The structure of the information is scoped to general-purpose scholarly documents, as opposed to for example encapsulating social (scholarly) Web activities.
• The structure and semantics of information is fundamentally based on HTML or XML (as opposed to the RDF language).
• In some cases, there is information duplicity as human and machine consumers are served with different data.
• There are post-processing dependencies (eg. via JavaScript execution) in order to turn hidden metadata into visible and readable by humans.

Research data in the form of observational or experimental data follows a different structure than narrative content – which is a common characteristic of a research article, annotation, notification or a profile. Structured data may be (semi-)automatically aggregated through instruments, stored, and viewed in different ways.

As pointed out in Statistical Linked Dataspaces, 2012, Capadisli, what linked statistics provide, and in fact enable, are queries across datasets: Given that the dimension concepts are interlinked, one can learn from a certain observation’s dimension value, and enable the automation of cross-dataset queries. Moreover, domain-specific user interfaces can be built based on federated queries as discussed in Linked Statistical Data Analysis in Sparqlines: SPARQL to Sparkline.

Enabling Scientific Data on the Web, Miłowski, 2014, focuses on enabling scientific data to exist on the Web in such a way that it can be processed both as viewable content and consumed data. Miłowski posits that in order to enable user interactions through the Open Web Platform while having the same information machine-readable, data must to be partitioned into “Web-sized” representations. This can be contrasted with the approaches mentioned from earlier where the data is treated as "as is" without being structured in any particular way for applications to consume.

In this thesis I describe and exemplify multiple implementation approaches. For example, the tabular data in Characteristics of Specifications, Comparison of Notification Mechanisms, Linked Data Notifications Implementations, and in Decentralised Linked Research Application, as well as the data in Linked Specifications, Test Suites, and Implementation Reports are all human- and machine-readable. Whereas in Interactive Linked Statistics, I will describe Sparqlines: SPARQL to Sparkline where SPARQL queries can be executed by client-side applications in an article authoring environment to fetch live data from remote endpoints, and be represented in SVG+RDFa.

Throughout history, the complex connections between communication technologies and society transformed control structures of civilisations. Harold Innis and Marshall McLuhan argued that electronic media, unlike any other, compressed time and space, creating a global village as well as facilitating the influence and potential of centralised control.

While the Internet reinforces democratic (re)distribution of communicative power, and the Web is inherently a decentralised social system, centralisation of all forms of information exchange and various forms of censorship – state or self-imposed – still occurs. In response to the Telecommunications Act of 1996, John Perry Barlow voiced a fundamental societal concern in A Declaration of the Independence of Cyberspace given the attempts of the governments of the industrial world to exert control over it, the builders of the cyberspace; a self-sovereign online society that is cross- border, culture, with diverse economies is ultimately an opposing force.

Over the years, large-scale centralised systems were built to collect, organise and share data in diverse sectors. The domination of centralisation is essentially due to economic and state incentives, and it has shaped the technical infrastructure: it is currently much easier and more efficient to author, manage, publish, read, and search large amounts of content using centralised services or proprietary platforms. As such the convenience that comes with centralisation is at the user’s expense where ownership and access control is compromised, and ultimately the user’s privacy and security. Moreover, from the service provider’s perspective, there is no particular incentive to build interoperable systems as that would entail losing control over data. Since such services hinder information flow; information in one service is not necessarily usable by other applications and services and vice versa, they effectively form information silos – Decentralization: The Future of Online Social Networking, 2008.

In addition to state controlled network infrastructures, information empires like Google, Facebook, Amazon, Elsevier, as well as a long tail of (social media) services, exert a great deal of influence on how information can be accessed, expressed and reused. This influence impacts society, affecting the cost of accessing information, the effort required to produce and distribute, long-standing privacy concerns (in exchange of some convenience for users), as well as how and the amount of information that is aggregated, shared and filtered.

Independence from centralised services and platforms is a necessity for ownership of published ideas, and to establish authenticity and trust. For example, Facebook has been accused of data scandals, bias, false information, and censorship – but rather than blaming this on any particular service or platform, we identify it as an unavoidable result of centralisation. After all, there is a continued tension between unrestricted publication rights on the one hand, and a guarantee of balanced, verified information on the other. In a fully decentralised setting, each source is filterless and responsible for its own quality and reputation, while consumers are free to selectively (dis-)trust certain sources using any mechanism they desire, for example through trusted parties. This can have both favourable and unfavourable consequences depending on actual implementations. Google Search or DuckDuckGo for instance apply different filtering algorithms based on their Web crawls. Whether consciously or not, why we use a particular search engine naturally correlates with the amount of use or trust we have on that system. Similarly, as the academic peer review system acts as a layer to regulate credibility of academic contributions, we can approach the research results provided that there is sufficient provenance, accountability of the quality control processes, and able to distinguish different levels of certification given or lack of from the scholarly community.

With the presumption that the major external forces – governments, companies – are at play, the situation for individual-controlled communication is complicated and ever challenging. Given that academic research is a global activity, it is important to acknowledge that the Web enables an environment in which actors in the space can both autonomously conduct their endeavours and participate in knowledge exchange. Hence, there is a need to re-decentralise the Web so that it benefits individuals, the public, as well as the scholarly communication market as a whole. We need to improve decentralisation and take further steps to apply it to research activities and scholarly publishing.

The information needs of individuals, communities and societies vary on personal, local and global levels. In Understanding Knowledge as a Commons, 2011, Hess and Ostrom, and contributing authors state that scientific communities can make their discoveries and digital resources available to all which can be shared without having a subtractable characteristic like that of print. Through electronic knowledge representations, researchers can take part in reproducing and replicating others’ findings; be involved in quality-control and certification processes, as well as incorporate interactions among researchers, examiners, journalists, policy makers, and the general public. Given access to the Internet and the Web, academic researchers can register their URIs, and control how they express their ideas and what they refer to, establish data storage and choose which applications to use for their social and scholarly activity.

So far I have used the term decentralisation liberally, but in fact it can mean different things depending on the context, community, and collection of standards and technologies that are used. From here on end, I work with the the following definition in context of the Web architecture:

Decentralized system

A distributed system in which multiple authorities control different components and no single authority is fully trusted by all others.

Messages and operations in centralised or distributed systems may be managed by a single root of trust or authority. From this position decentralized systems are a subset of distributed systems where they embody a complex set of relationships of trust between parties managing different aspects of the system. These systems are conceived of as a graph where the nodes correspond to the components of the system, and the edges correspond to connections between them. From a global perspective, the Web is a decentralised system, however, it can have components which are disjoint from the rest of the graph. Figure Coupling of Centralised and Decentralised Architectures depicts the contrast between typical centralised and decentralised architectures.

Troncoso et al, consider security and privacy to be integral aspects of systems. Security includes confidentiality, integrity, authentication … availability, accountability, authorization, non-repudiation or non-equivocation, and privacy considers protections of users’ related data (identities, actions, etc.), and formalized in terms of privacy properties (anonymity, pseudonymity, unlinkability, unobservability).

How do we control ourselves on the modern, overly-centralised Web? I focus on the feasibility of using a set of open Web standards and technologies to address certain kinds of challenges in scholarly communication. I conclude with a focus on three areas of control (Degree of Control) that are at the forefront of the way (scholarly) communication is conducted on the Web and its ramifications on the rest of the sociotechnical system.

In this section I discuss existing research involving the decentralisation of dataspaces on the Web.

In From Databases to Dataspaces: A New Abstraction for Information Management, Franklin, 2005, raises a challenge of organisations relying on large number of diverse, interrelated data sources, but having no way to manage their dataspaces in a convenient, or principles fashion.. The article brings forth the notion of dataspaces as a data co-existence approach, with the goal of providing functionality over all data source. In such systems, there is a set of participants and relationships, while remaining sensitive to their requirements of autonomy. The authors propose the development of different DataSpace Support Platforms (DSSP) where interrelated services would allow data to be managed by participant systems.

In Principles of Dataspace Systems, Halevy, 2006, explores challenges that are involved in realizing DSSPs based on motivating applications eg. personal information management (PIM), scientific data management, structured queries and content on the Web. Authors conclude that interoperability and interchangeability of multiple components is key to the success of DSSPs.

In Web-scale Data Integration: You can only afford to Pay As You Go, Madhavan, 2007, propose the notion of a PAYGO data integration architecture with the premise that it is impossible to fully integrate vast heterogeneous collections of structured data that exists on the Web. The PAYGO principle states that a system needs to be able to incrementally evolve in its understanding of underlying data’s structure, semantics, and relationships between sources.

In A decentralized architecture for consolidating personal information ecosystems: The WebBox, Van Kleek, 2012, proposes a realisation of Berners-Lee’s socially aware cloud storage, in which Web applications running on a user’s devices gain privileged access a unified, user-controlled data space, where Webbox applications are decoupled from personal data spaces.

In Data ownership and interoperability for a decentralized social semantic web, Sambra, 2013, acknowledges the same core challenges of the modern Web as Tramp, and sets out to identify key components that would help achieve data ownership and interoperability. Sambra demonstrates how a stack of interoperable Web technologies around WebID and Linked Data can be used to build class of social Web applications, where users without centralised intermediaries can authenticate themselves, and participate in creating and exchanging data with servers that are equipped with social access controls.

In Amber: Decoupling User Data from Web Applications, Chajed, 2015, posit that users control their own data choosing which applications to manipulate their data as well as their ability to share data between applications and with other users. Amber is a proposed architecture that looks into overcoming challenges on inexpensive querying, an expressive authentication and access control system, users trusting their platform, being able to handle large volumes of data and query results, offline capabilities, and a sensible economic compatibility among its users, platform, and application developers. The results from Amber show that the architecture is most useful when data is associated with users, and operations on the data are carried out in Web browsers.

Standards-based Scholarly Communication: In order to manifest a Web-centric interoperable information space, I investigate existing research, Web specifications and practices that can fulfill the forces and functions in scholarly communication. This is in contrast to an arbitrary ensemble of implementations which are not guaranteed to be interoperable with each other, alternatively may only after post-facto interoperability initiatives eg. the survey on 400+ Tools and innovations in scholarly communication for the most part observes applications implementing propriety designs. The following are a selection of open Web specifications; standards and practices that are designed with the notion of decentralisation, interoperability, and extensibility. They are based on use cases that facilitate discovery, read-write operations of interlinked Web resources.

Both Tramp and Sambra approach from the perspective that the Web has proven itself architecturally and universality, and argue to extend the decentralised architecture for information reuse and user interaction. The core approaches use the Linked Data technology stack for publishing, retrieval, and integration; decoupling services and applications from the users data so that content creators are owners and have desired rights on their data and its use; and fostering data extensibility and distribution, as well as privacy.

The foundational communication and transfer protocols of the Internet and the Web are designed to operate in a decentralised and distributed manner. In terms of global functionality, removing a single node from the network – a machine with an IP address – does not halt the whole network. Similarly, when an HTTP URL is no longer available (eg. typical "404 Not Found", "410 Gone" status messages), the other URLs are not directly affected. This particular behaviour, where the Web being functional without requiring bi-directional hyperlinks is considered to be a "feature" (rather than a "bug") of the overall system. In the meantime, bi-directional links can still be established when so desired eg. as in the case with notifications.

The AWWW enables the identification, discovery and description of information for the notion of a Read-Write Web where servers and clients interact using Web standards.

Linked Data Platform (LDP) specifies a RESTful read-write protocol and vocabulary for Web resources. LDP has distinct notions for resource and container; where the container construct can group and manage resources. LDP uses the RDF data model to describe the state of RDF and non-RDF resources. Linked Data Platform Use Cases and Requirements outlines a wide-range of user stories, use cases, scenarios, and requirements which was used as the bases for the LDP specification. LDP can be used to mimic a file system abstraction interacting with Web resources over HTTP. The LDP Paging mechanism can be used by clients and servers to efficiently request and serve a resource description, eg. resources in a container, in multiple parts. All LDP servers and clients can interact with each other using the same HTTP interface in context of Linked Data operations. An LDP server can facilitate researchers to register and store their resources, as well as share their content by applying different access controls.

Hydra is a vocabulary to enable the creation of hypermedia-driven Web APIs by enabling a server to advertise affordances of its resources – machine-readable valid state transitions – to a client. A client uses this information to construct HTTP requests in order to perform possible read-write operations. Hydra enables generic client applications to be built without hardcoding knowledge about the available operations against Web APIs. With Hydra, applications can automate some of their tasks and only request researchers’ engagement when need to.

Linked Data Fragments (LDF) is the concept of having a uniform view on Linked Data interfaces that can be used towards reliable Web querying. To that end, a client asks a server about the kinds of Linked Data fragments that are available and then dynamically adapts its query plan for the server to execute. A Linked Data fragment is defined by three characteristics: data (what triples does it contain?), metadata (what do we know abut it?), controls (how to access more data?). LDF-based interactions can help researchers to inspect available scientific data through their distinct characteristics.

Triple Pattern Fragments (TPF) is one possible way for a server to define a Linked Data fragment that enable live querying over the dataset on the client-side. TPF’s hypermedia controls are expressed using the Hydra vocabulary.

Linked Data Templates (LDT) provides the means to define read-write Linked Data APIs declaratively using SPARQL and specify a uniform interaction protocol for them. An LDT based server applies re-usable RDF ontologies to define application structure declaratively as a set of instructions for resource representation processing. To this end, templating is where an HTTP operation is mapped to a certain URI pattern that a server executes a SPARQL command in order to drive RDF processing. When a client triggers a particular request pattern, the server gives a suitable response based on available templates that are identified with URIs. LDT allows arbitrary academic and research applications to be built while exposing a uniform API that can be used in decentralised environments.

RDF/POST Encoding for RDF uses the W3C HTML 4.01 Recommendation to specify read-write operations of RDF data through form-encoded RDF serialisation. Read operations are done by having an RDF/POST document encoded in HTTP GET URL, and write operations are sent via HTTP POST with form-urlencoded media type. Operations similar to LDP and LDT can be achieved with RDF/POST.

Fedora API Specification refines the semantics and interaction patterns of the LDP specification in order to address the needs of repositories for durable access to digital data. Its goal is to facilitate interoperability with client applications. While the interactions patterns provides a mechanism for different client applications to communicate with servers, servers can be expected to vary in the kinds of services and affordances they offer for their resources. Fedora uses LDP as a foundation and defines the version identification and navigation scheme with respect to the Memento specification, integrates resource authorization and Access Control List rules (both described below), provides a design for the publication of event notifications, and interaction patterns to support binary resource fixity verification. The Fedora API can fulfill the registration, awareness and archiving functions, while enabling different actors to cooperate across instances on each others resources through independently built applications.

HTTP Framework for Time-Based Access to Resource States -- Memento (RFC 7089): introduces a uniform, datetime-based, version access capability that integrates present and past Web resources. As representations of resources change over time, there is a need to preserve (archive and version) earlier representations. One common resource versioning pattern on the Web consists of generic URIs referring to the latest version of an accessible resource, as well as having a dedicated version URI for each resource. The Memento framework facilitates the discovery and retrieval of distributed versioned resources with datetime negotiation, a variant of content-negotiation; and TimeMaps an index of URIs referring to the prior states of a resource. The framework also helps to recognise frozen states of resources – a promise that the resource state has not and will not change. The Memento protocol can assist scientific applications to discover and reveal the variations of scholarly artifacts.

Web Annotation (WA) is set of specifications for an interoperable, sharable, distributed Web Annotation architecture. The annotations convey information about a resource or associations between resources to meet different motivations and purposes, eg. assessing, replying, describing, bookmarking, as well as linking arbitrary content to a particular data point or to segments of timed multimedia resources. Web Annotation Protocol (WAP) is the HTTP API for publishing, syndicating, and distributing Web Annotations. Much of the protocol is based on using and extending LDP and REST best practices. Web Annotation Data Model describes the underlying annotation abstract data model as well as a JSON-LD serialization. The WAP uses an LDP container to manage annotations with some constraints derived from the WA Model. Scholars can use servers and applications that support Web Annotation in order to annotate scholarly literature to meet the certification function.

Activity Streams 2.0 (AS2) detail a model for representing potential and completed activities with the intention of using specific types of activity vocabularies defined elsewhere. The Activity Vocabulary provides a foundational vocabulary to describe past, present, and future activities eg. announcing, creating, following, offering, and about objects eg. actors, media. Various academic and research activities can be recorded and brought to the attention of actors to determine their applicability and allow management of scholarly knowledge.

Linked Data Notifications (LDN) is a resource-centric communication protocol for applications to generate notifications about activities, interactions, and new information, which may be presented to the user or processed further. It allows any resource (target) to advertise a receiving endpoint (inbox) for the messages anywhere on the Web. The server (receiver) hosting the inbox can have messages pushed to them by applications (sender), as well as how other applications (consumer) may retrieve those messages. Each notification in an inbox is an identifiable and reusable unit (URI), and can contain any data using any vocabulary. LDN can be combined with Web Annotation and Activity Streams to support applications to deliver and consume notifications about research and scholarly activities. I will revisit LDN and discuss its role in research communication where it supports the accessibility, content forces, and how it can perform the registration and awareness functions.

ActivityPub (AP) is a decentralised social networking protocol based upon the Activity Streams 2.0 data format. It provides a client to server API, as well as a federated server to server API for delivering notifications and content. User accounts on servers have an inbox (to receive messages from the world), and an outbox (to send messages to others). AP’s inbox property is the same as LDN’s, and the targeting and delivery mechanism can be interoperably combined. AP can facilitate distributed scholarly interactions by allowing actors to disseminate their operations on their own, as well as others’ data. It serves to fulfill various social scholarly Web functions.

The design pattern for these read-write Linked Data centric protocols and architectures can be categorised as follows:

• The ability to perform HTTP interactions against Web resources.
• Decoupling software from domain or application-specific operations and data.
• Descriptive, extensible, and interconnected units of information.
• Declarative and machine-readable affordances for valid read-write operations.

One of the functions of an online identity is to a identify a persisting entity at a particular time and context. The construction of an online identity of an entity eg. individual, organisation, community, state, can be complex in that it is interconnected with social and technical systems, and it can be issued – declared to exist, named, or be referenceable – in different ways. Entities being universally and uniquely identifiable on or off the Web entails that they can be attributed for their contributions, develop their reputations, be accountable for their actions and research output, as well as play an integral role towards building trust. An actor is an indispensable force in research communication, and enabling identity agency is both relevant and foundational.

In The Presentation of Self on a Decentralised Web, Guy, 2017, the What is a profile? study describes affordances of systems which integrate online profiles in a social capacity and raises five features of systems with regards to their representations of users: flexibility, access control, prominence, portability, representation. Furthermore, Guy posits that online self-presentation is both constituted and affected by who sees a representation of an individual, and what it is they see, both of which are encompassed by the situation whereby it is seen. The results of the studies are summarised as a conceptual framework as the 5Cs for online profiles and self presentation: context, control, customisability, connectivity, cascade.

Third-party controlled identity: Community or state controlled identifiers for identities are a class of identifiers that are ultimately third-party controlled in that their ownership, governance, description is beyond an individual’s jurisdiction. The creation and control of identifiers by a group has the quality of long-term promise for persistence. Typically there are policies in place which promise their longevity, and procedures in the event that the group ceases to manage the identifiers. For example, ORCID URIs are intended to be part of the scholarly commons where they managed by the ORCID community. The ORCID URI and profile description is useful to researchers, organisations, as well as to interlink researcher objects, however it is ultimately intended to serve the research community, and the systems that use it. One unintended consequence of this kind of centralisation in scholarly communication is that a growing number of institutions and publishers require researchers to participate with their ORCID identifiers, as opposed to any URI of a researcher. The resulting effects of such initiatives are counter to autonomous participation in that a researcher has to conform to the terms and policies of multiple third-parties. On the other hand, the administration of the URI by the ORCID organisation removes the burden of maintaining the URI as well as the profile URL. In this case, a "trusted" third-party takes care of the identifier, whereas researchers are responsible of updating the content.

While some scholarly societies adopt identifiers like ORCID given their agreed potential benefits, it can also be harmful in the case of tracking researchers; where they work, who they work for or collaborate with, what they work on. Such ease of identification opens up the possibility to target and surveil researchers, as well as to discriminate or exile them if their findings act counter to others’ interests. [GOAL] eLife collects ORCIDs from authors of accepted papers at proofing, Morrison, 2017, argues that it is not clear that facilitating researcher identification is in the best interests of academic freedom … participation in a service like ORCID should be optional. User Tracking on Academic Publisher Platforms, Hanson, 2019, studied 15 different publisher sites and found that on average, each had 18 third-party assets loaded, with a total of 139 distinct third-parties across the platforms. The loaded JavaScript and cookies tracked user behaviour, interests, and research when combined with user identity information or fragmentary identity. The collected user data is shared with ad networks and data brokers. Unauthorised third-party tracking of users online activities is particularly concerning even in the case of public sector websites, which are expected to be more privacy respecting than "ordinary" websites (as per GDPR in the EU). For example, Ad Tech Surveillance on the Public Sector Web, Cookiebot, 2019, reports that advertising technology (ad tech) companies are extensively tracking EU citizens using government and public health service websites – a total of 112 companies were identified using trackers that send data to a total of 131 third party tracking domains. Thus, behavioural data combined with third-party identifiers in any shape or form (including syncing across services) is in breach of individual’s privacy in many jurisdictions.

Self-controlled identity: The concept of a personal self-sovereign identity entails that entities can register, describe, and manage their own identities of their choice. Controlling our online identities is important because it reflects our individuality and privacy. When identities are verifiable, we can also distinctly identify applications and their (human or machine) users, as well as switch between them depending on the context of use.

Different types of strings, eg. a username, an email address, an RSA public key, a URI, can be used to directly or indirectly identify agents. While each can be useful within their own context, Web-based identifiers (HTTP URI) make it possible to provide machine-readable descriptions when dereferenced, descriptions being extensible by their owners, as well as being interoperable global identifiers that can be interlinked unambiguously with other things.

A server can collect information about a client based on its device, machine, or browser fingerprint, even if the user did not directly disclose information about themselves. On the other hand, global identifiers like government-issued identifiers, credit card numbers, and full names, can be used to determine, track, and correlate an entity.

One of the conclusions of the workshop was that many of the existing technologies needed to create decentralised social networks already existed, eg. FOAF, OpenID, XMPP. One of contributions to the workshop was presented by Henry Story based on the original proposal for FOAF+SSL (combining RDFAuth by Henry Story and sketch of a simple authentication protocol by Toby Inkster in 2008) for the notion of using Semantic Web vocabularies such as FOAF with SSL certificate exchange mechanism permits distributed and interlinked social networks to exist. The protocol was later posited in FOAF+SSL: RESTful Authentication for the Social Web. Its goal was set to protecting and controlling access to personal information distributed on the Web where identification and privacy is at its centre. The work on WebID and related extensions described below builds on this work.

In order to enable people, group and software to act as autonomous agents that can cooperate on the Web, we need methods for distributed and decentralised identity, secure authentication, and access control. The collection of WebID specifications outlines the approaches to meet those needs.

Web Identity and Discovery (WebID) is a specification that outlines a distributed and extensible universal identification mechanism on the Web. A WebID is an HTTP URI denoting an agent, for example a person, organisation, or software. It can be used towards the declaration of an agent’s existence; discovery, describing or reference; authentication; and authorization. WebID is distributed and decentralised in that any agent can be named anywhere on the Web, and the server hosting the URI space controls the identity. It is extensible in that it can be used in conjunction with secure authentication and access control mechanism, as well as having dereferenceable machine-readable descriptions which are themselves extensible. WebID can improve privacy, security, and control be in the hands of users. WebID fulfils registration of identities of academics and researchers, helps them to be associated with scholarly artifacts they interact with eg. authoring an article, having access to an annotation.

WebID Profile is an RDF document which uniquely describes an agents denoted by a WebID. The profile description is a way for researchers to extend the information about themselves; specify their preferences for applications, the location of their storage, and scholarly artifacts; how they can be contacted; their associations with other researchers, as well as other identities, and so forth. It serves to make identity claims which can be coupled with verification systems.

I have authority of the domain name csarven.ca where I publish my WebID Profile document https://csarven.ca/ describing my WebID https://csarven.ca/#i referring to me (in real life).

<https://csarven.ca/>
  a foaf:PersonalProfileDocument ;
  foaf:primaryTopic <https://csarven.ca/#i> .

<https://csarven.ca/#i>
  a foaf:Person ;
  foaf:name "Sarven Capadisli"@en ;
  foaf:mbox <mailto:info@csarven.ca> ;
  foaf:img <https://csarven.ca/media/images/sarven-capadisli.jpg> ;
  foaf:interest <https://en.wikipedia.org/wiki/Media_theory> ;
  cert:key <https://csarven.ca/#key-e785> ;
  foaf:account <irc://irc.freenode.net/csarven,isnick> ;
  foaf:knows <https://www.w3.org/People/Berners-Lee/card#i> ;
  pim:storage <https://csarven.ca/> ;
  ldp:inbox <https://csarven.ca/inbox/> ;
  as:outbox <https://csarven.ca/outbox/> ;
  foaf:made <https://dokie.li/> ;
  rdfs:seeAlso <https://csarven.ca/cv> .

Multiple identities: Individuals may want to be associated with multiple identities to fulfill diverse needs and social expectations. For instance, academics, politicians, employees, and artists, to name a few, may want to perform their online activities using public identifiers, so that they can be discovered, attributed, rewarded, trusted etc. There can also be quasi-public identities issued by a government or a bank to connect individuals to the physical world. Furthermore, individuals may wish to have several personal – public or private – identities and self representations, eg. on social networks, to address different needs. At any event, identities and corresponding identifiers can be registered and managed by different entities, and may be interlinked.

For instance, while ORCID profile description has constraints, it allows extensibility in the form of asserting a "same as" relation through its UI, so that the ORCID URI can be linked to other WebIDs. This essentially makes a connection to another self-describing document and facilitates follow your nose discovery. My personal WebID which is under my control is https://csarven.ca/#i. I also have an ORCID identifier that also serves as a WebID: https://orcid.org/0000-0002-0880-9125 which also has an RDF representation. As ORCID’s user interface allows people to extend their descriptions with certain fields, it is possible to add same as or preferred URI relations:

<https://orcid.org/0000-0002-0880-9125> ;
  owl:sameAs <https://csarven.ca/#i> ;
  contact:preferredURI "https://csarven.ca/#i" .

The owl:sameAs and contact:preferredURI information provides an opportunity for applications consuming the ORCID WebID to potentially discover more information about other entities. This is particularly useful as the ORCID WebID Profile is not completely open to add arbitrary information. For example, specifying a person’s picture or their contacts is not currently possible with ORCID, however, if the alternative profiles (https://csarven.ca/#i) contain such information, they can be purposed to know more about the entity.

• identifying a layer of trusted (or acknowledged) network of people
• sending notifications to the members of one’s personal social graph
• discovering annotations made by a particular social graph

Both the entities (nodes) and the relations (edges) of a social graph may have public and private parts. An example of a social graph where someone knows someone that knows someone, is as follows:

<https://www.w3.org/People/Berners-Lee/card#i>
  foaf:knows <https://csarven.ca/#i> .

<https://csarven.ca/#i>
  foaf:knows <https://bblfish.net/people/henry/card#me> .

Trust is ultimately tied to verifiable claims about identity and their provenance. While the creation and protection of third-party controlled online identities have received much focus, claiming and using personal identities online is still a challenge. Today’s Web applications predominantly make use of online identities that are essentially tied to and controlled by authorities in which users are merely borrowing an identifier and its management. It is a common practice to where users are required to create accounts at each online service they wish to use, eg. social media sites. Moreover, the trust model of such design dictates claims and usage of online identities is not ultimately in the hands of the users, and can be taken away from the users by the owners of the URIs and service at any time and reason. Even in cases where third-party issues identities can be used across different services, it is subject to a privacy violation as the identity provider essentially gets to know which services the user’s online identity is used with and for what purpose. Consequently such services can gradually build up information about users’ actions as well as behavioural patterns on the Web. Lastly, portability and interoperability of the identities is generally under developed as there is no particular incentive for an identity provider to make it so – after all, providing “free” services for online identities, and then locking identities to well-defined circle of services is taken advantage of by large corporations given their business models. Perhaps most importantly, such online services (as well as applications) do not recognise the identity of users as a distinct notion from the applications and services they use towards authentication.

While there are different authentication mechanisms, here I limit our coverage based on open standards and decoupling of identity, authentication and authorization process, as well as the resource server that the accounts are available from. Hence I categorically omit service-based authentication workflows which are known to require an account name and password, as well as for instance The 'Basic' HTTP Authentication Scheme (RFC 7617).

Web of Trust (WoT) is a concept for a decentralised trust model used in systems like Pretty Good Privacy (PGP) to establish the authenticity of the binding between a public key and its owner. A public key cryptographic system uses a pair of keys: public keys (disseminated openly) and private keys (withheld by the owner). It has two main functions: authentication and encryption. For authentication, public key verifies that a holder of the paired private key sent the message. In the case of encryption, for example, when Guinan wants to send Picard a message, Guinan takes Picard’s public key to encrypt a message where only Picard can decrypt the message with his private key. WoT essentially leaves trust decisions to the users, as opposed to a centralised authority. Thus, WoT uses self-signed certificates and third-party attestation of those certificates, eg. at key signing parties. Self-signing certificates removes the dependency on hierarchical certificate authorities to assert identity.

X.509 is a standard for public key certificates that specifies a structure for a digital certificate, including a public key associated with one or more identities, as well as extensions for new information. X.509 essentially serves as a identity card that can be used to store identity claims. X.509 certificates can be used to authenticate both servers and clients. HTTP over TLS (RFC 2818) – an extension of HTTP for secure communication (HTTPS) – uses the Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile (RFC 5280). All combined, HTTPS ensures privacy, integrity, and identification.

The Transport Layer Security Protocol Version 1.2 (RFC 5246) is a cryptographic protocol aimed at providing secure connections between applications. When applications eg. a server and a client (Web browser), communicate using the TLS protocol, it is private (secure) in that data is encrypted, identity of either party can be authenticated using public key cryptography (like PGP), and integrity of the exchanged data can be assured during transit.

WebID-TLS defines how a server can authenticate a user with a client application holding the user’s public key certificate, and serves towards identity claim verification. As the extension mechanism in X.509 allows additional statements to be specified under the Subject Alternative Name (SAN) field, it is used to declare the user’s WebID for the URI field. When a client application running on the Web browser requests a resource from the server, the TLS connection gets established between them. Once TLS has been set up, the HTTP application protocol exchange takes place (eg. HTTP GET, HTTP POST). If the requested resource requires WebID authentication the server can request the client to authenticate itself. When the Web browser encounters this certificate request, it may prompt its user to choose a certificate (from its certificate manager) – thereby signing a token with its private key – to let the client send it to the server. The verification agent extracts the public key as well as the WebID from the certificate. As the WebID enables a global dereferencing mechanism for finding a key, the server uses can decide on dereferencing the WebID Profile and verify the claims about the WebID. Once the identity verification is established, the server can look up its access control rules for the requested resource to determine if and how the request can be fulfilled. Researchers can use servers and applications implementing the WebID-TLS workflow towards securely identifying profile claims and to have access to resources may be otherwise protected.

The Cert Ontology 1.0 can be used in WebID Profile documents to specify an agent’s certificate information. For example, an agent can indicate their public and private keys. The fields and corresponding values of Public Key Algorithm, Modulus, and Exponent in the X.509 certificate are used in the WebID Profile as part of cert:RSAPublicKey, cert:exponent, and cert:modulus. URI under SAN is our WebID. An example where information about an RSA Public Key is associated with a WebID:

<https://csarven.ca/#i>
  cert:key <https://csarven.ca/#key-e785> .

<https://csarven.ca/#key-e785>
  a cert:RSAPublicKey ;
  cert:exponent "65537"^^xsd:nonNegativeInteger ;
  cert:modulus "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"^^xsd:hexBinary .

WebID Profiles using the Cert Ontology can help authenticating servers to discover researchers public keys which can be used towards verifying identity claims.

WebID-TLS-Delegation is an an extension to WebID-TLS that enables users to delegate other agents to act on their behalf. In Extending the WebID Protocol with Access Delegation, Tramp, 2012, states how a principal agent can delegate authentication and access to a trusted secretary agent – typically a third-party software – who can act on behalf of of the principal agent to perform (asynchronous) requests. A secretary with its own WebID acting on behalf of a user can potentially have the same (or required) privileges while still being distinguishable eg. via different public key, and capabilities. An agent can declare a delegation as follows:

<https://csarven.ca/#i>
  acl:delegates <https://example.org/application#i> .

whereas the agent making requests on behalf of another agent would include the HTTP On-Behalf-Of header when making requests.

OpenID Connect (OIDC) is an authentication extension to The OAuth 2.0 Authorization Framework (RFC 6749). OAuth 2.0 provides an authorization layer that separates the role of a client from that of the resource owner. Clients obtain an access token issued by an authorization server with the approval of a resource owner, and clients can use that delegation-specific credentials (access token) to obtain and use limited access to HTTP resources on the server. In addition to the authentication process, OIDC enables authorization servers to obtain profile information about the user.

WebID-OIDC is an authentication delegation protocol resulting in OIDC’s verified ID Token once a client authenticated, and deriving the WebID from the ID Token. The complete WebID-OIDC workflow as follows: an initial request is made to access a protected resource; whereby the user specifies their identity provider to the authorization server; the user authenticates themselves at the provider and gets redirected back to the resource they were trying to access and provides the signed ID Token; the server controlling the resource validates the token and extracts the WebID; and finally the confirmation from the resource server. For researchers, WebID-OIDC is equally applicable as WebID-TLS.

Access Control List (ACL) ontology, enables a server to provide four modes of access: Read, Write, Append, and Control. As implied by the name, read allows a server’s resources to be accessible and interpretable. Write allows resources to be modified or deleted. Append allows a specific kind of Write in that while adding of information is permitted, removal is not. Control permits full read and write access to an ACL resource, which is typically given only to the "owners" (or administrators).

The following code snippets of authorization policies are used to specify the ACL information (eg. resource.acl) associated with a resource (eg. resource) on a server. For example, setting an authorization policy where an agent (https://csarven.ca/#i) has write access to resources, as well as control access to their respective ACLs:

<#write-control>
  a acl:Authorization ;
  acl:agent <https://csarven.ca/#i> ;
  acl:accessTo <./> ;
  acl:mode acl:Write, acl:Control .

Setting a class of agents to perform read and append operations on a resource eg. an inbox:

<#read-append>
  a acl:Authorization ;
  acl:agentClass foaf:Agent ;
  acl:accessTo <./> ;
  acl:mode acl:Read , acl:Append .

ACL plays a role in determining which authorization policies to apply when a resource is requested eg. raw sensitive data, private social graph, initial versions of an article.

Prohibiting Delegation, Miller, 1999, provides an analysis on four delegation problems: perimeter security, confinement distributed confinement, confused deputy, communicating conspirators, and examines the basic differences between Capabilities and ACLs. The article posits that users and programmers can be lead to make decisions under a false sense of security about what others can be prevented from doing, ACLs seduce them into actions that compromise their own security.

As knowledge builds on knowledge, it is vital to preserve the connection between units of information in scholarly communication. There are a number of challenges in this respect. In this section I focus on the persistence of the identifiers for units, as well as content integrity and preservation of context. In the most general sense, trust and accessibility are integral to preservation of content, integrity, and the context in which they are used. This is so that knowledge can be scrutinised with the help of transparent trail of reproducibility and replicability of research results. I look at existing research and approaches in this space.

Domain name registration and maintenance is one of the key factors for long-term reliable persistence as per URI ownership. Berners-Lee outlines two issues for the persistence of HTTP URIs:

1. The persistence of the opaque string which follows the domain name, and
2. the persistence of the domain name itself.

In Cool URIs don't change, Berners-Lee, 1998, discusses some of the approaches that can be taken towards usefulness and longevity of URIs. The article focuses on practices that a publisher making a commitment to persistence by designing and managing the URI path and the content it resolves to, as well as the domain name it uses. The owner of a domain name has the obligation to define what the things mean for its URIs. This is also a form of social contract made by the authority that names and defines a URI to anyone that’s using it – see also Philosophical Engineering and Ownerhip of URIs.

Persistence policies can come in different forms. For example, W3C’s URI Persistence Policy is a document making a pledge about how some of the resources under its domain will persist throughout the lifetime of the Consortium; any changes to persistent resources will be archived; and in case that the organisation is disbanded, its resources can be made available under the same rights and license. These human-readable statements are useful institutional commitment to persistence. The ODRL vocabulary can be used in a similar way to provide a machine-readable policy about resources.

Decentralized Identifiers (DID) are identifiers for verifiable self-sovereign digital identity, where they are under the control of the subject, independent from any centralized registry, identity provider, or certificate authority. In a way, DIDs go around the shortcomings of the domain name system where they can be created and managed without the authority of the registrar.

In Analyzing the Persistence of Referenced Web Resources with Memento, Sanderson, 2011, presents the results of a study on the persistence and availability of Web resources cited from research articles in two scholarly repositories. The results show that within a few years of the URL being cited, 45% of the URLs referenced from arXiv still exist but are not preserved, and 28% of the resources referenced by articles in the UNT digital library have been lost. In order to address this commonly known as URIs ceasing to exist (link rot), authors suggest that repositories expose the links in the articles through an API so that Web crawlers can be used to archive. With the help of archives supporting the Memento protocol, the original context of the citation can still be reconstructed.

Given the dynamic and ephemeral nature of the Web, and in particular management of URIs and corresponding representations at URLs, it poses a threat to integrity of Web-based scholarly content, and the consistency of scholarly records, as well as everywhere else. One special area is about the formal citation of scholarly resources eg. DOI, HTTP-DOI-URI, and informal referencing of other resources on the Web ie. any HTTP URI. The Hiberlink project investigates reference rot in Web-based scholarly communication, and introduces the term to denote two problems in using URI references:

Link rot

The resource identified by a URI may cease to exist and hence a URI reference to that resource will no longer provide access to referenced content.

Content drift

The resource identified by a URI may change over time and hence, the content at the end of the URI may evolve, even to such an extent that it ceases to be representative of the content that was originally referenced.

In Scholarly Context Not Found: One in Five Articles Suffers from Reference Rot, Klein, 2014, acknowledge extensively studied phenomenon on link rot and content drift, and themselves investigate the extent to which reference rot impacts the ability to revisit the web context that surrounds STM articles some time after their publication. The results show that significant amount of HTTP URIs cited in STM articles are no longer responsive or adequate archived snapshots available. Authors state that it is impossible to adequately recreate the temporal context of the scholarly discourse, hence suggest that robust solutions are needed to combat the problem of reference rot. Authors can take practical steps to remedy some of these issues eg. using archives that support on-demand snapshots, embedding the archived URI with datetime information alongside the reference to the original resource – I discuss this further in Robust Links.

In Scholarly Context Adrift: Three out of Four URI References Lead to Changed Content, Jones, 2016, reuse the same dataset from Klein study, to investigate to what extent the textual content remained stable since the publication of the referencing article based on various well-established similarity measures based on the comparison of the representative Memento and the live resource. They find that for over 75% of references the content has drifted away from what it was when referenced. The authors support the idea that in order to partly work around this issue, authors should pro-actively create snapshots of the referenced resources at Web archives, and referencing them in their scholarly literature. However, the authors also state that such robust embedding in the infrastructure of the existing authoring, reviewing, and publishing workflow is still an open challenge. To that end, applying the Robust Links approach can help. While the DOI-paradigm for scholarly units help to improve the link rot scenario when the custodians of the domains or the URLs of the scholarly resources relocate, the resources on the Web at large remain to be a problem given that their incentives towards longevity and access differ.

From the point of persistence of the domain name (losing ownership) as Berners-Lee describes, one kind of content drift would be if the content published at https://csarven.ca/ today may be different tomorrow if another authority gets to own csarven.ca and defines what goes there. Alternatively, it may be that the content at that location is dynamic, and could differ from one request to another. In both cases, content drift creates a situation where if the originally referenced resource is still the same. From the perspective of Web-based scholarly publications, changes to content – accidental or intentional – can impact the degree of reproducibility, replicability, comparability of research results, and maintaining a reliable scholarly record.

In Persistent Identifiers for Scholarly Assets and the Web: The Need for an Unambiguous Mapping Van de Sompel, 2014, posit that while PIDs are assigned to resources outside of information access protocols (like HTTP), there is a need to unambiguously bridge the discovery of the Web-oriented resource eg. from PID to HTTP URI, in a way that is machine-actionable. For example, the PID paradigm has the following discovery path:

1. PID is the resource identifier eg. 10.2218/ijdc.v9i1.320
2. HTTP-URI-PID is the resolving URI eg. https://doi.org/10.2218/ijdc.v9i1.320
3. HTTP-URI-LAND is the redirect URI (landing page) eg. http://www.ijdc.net/article/view/9.1.331
4. HTTP-URI-LOC is the location URI of the content eg. http://www.ijdc.net/article/download/9.1.331/362/

Authors propose that using existing standards and practice, the essential ingredients for such a mapping is as follows. A PID has a Web equivalent HTTP-URI-PID, which is a requisite – minted by the naming authority. The HTTP-URI-PID can be content-negotiated to result in a) a human-readable representation at HTTP-URI-LAND or b) machine-readable representations with distinct HTTP-URI-MACH. The HTTP-URI-LAND remains the same for discovery, however, HTTP-URI-MACH uses an RDF-based approach to describe the aggregations of scholarly assets based on the OAI-ORE specification.

It is worth briefly revisiting the notion of social agreements around Web resources. Conceptually, the agreement that I make with you about the persistence of my website’s resources is in essence the same as a naming authority controlling a PID, as well as all of the nodes in between HTTP-URI-LOC. The kind of mapping between a domain name and the IP address it points to is similar to a PID being mapped to a HTTP-URI-PID. Hence, URI Ownership essentially involves two possibilities: either I “own” and control a URI space or someone else does.

In Persistent URIs Must Be Used To Be Persistent, Van de Sompel, 2016, reveal the results of a study where authors do not use persistent URIs like DOIs even when available, and instead use the location URIs. In order to alleviate this issue, authors propose that an HTTP Link header is used at the location URI to announce the identifying HTTP-URI-PID. The current proposal is to use cite-as: A Link Relation to Convey a Preferred URI for Referencing, Van de Sompel, 2019, and a number of related patterns are outlined at Signposting the Scholarly Web.

Robust Links, Van de Sompel, 2015, is based on the idea that in a long enough time line, content-drift and link-rot pose challenges to maintain context and integrity between resources. In order to preserve the context of linked resources at the time of linking, a link robustness strategy is applied by way of creating a snapshot of the original resource and making it available to data consumers. The Link Decoration approach decorates links in hypertext documents in order to be human- and machine-actionable. By additionally associating the datetime information when linking helps to retain such context. The datetime information can be accompanied with a snapshot of the target URI made at a Web archive or a versioning system when linking, and using either the original or the versioned resource as the primary reference. Including datetime information in the source document can make it possible to interpret the context of each of the links in the document and to discover relevant snapshots. In the event that a linked resource or a snapshot becomes inaccessible, the datetime information helps to discover temporally close alternative snapshots.

Trusty URI is a technique to include cryptographic hash values in URIs to uniquely associate them with an artifact. In Trusty URIs: Verifiable, Immutable, and Permanent Digital Artifacts for Linked Data, Kuhn, 2014, outline how specific resources and their entire reference trees can be verifiable. If the trusty URI of an artifact is known, it can be used to verify if the content of an artifact corresponds to what it is suppose to represent. This is useful to determine if the content is corrupted or manipulated. It follows that trusty URI artifacts are immutable as each version of a content generates a unique trusty URI. Trusty URI artifacts are considered to be permanent in that once archived or cached, the artifact can still be verified even if the original location is no longer available. Applications can use trusty URI to encode their own references, as well as compute an artifacts’ trusty URI and verify before using it. Trusty URIs can be used to represent byte-level file content and for RDF graphs, and is compatible with named information URIs – Naming Things with Hashes (RFC 6920). An example from the Trusty URI Specification – Version 1: given resource http://example.org/r1, its trusty URI would be http://example.org/r1.RAcbjcRIQozo2wBMq4WcCYkFAjRz0AX-Ux3PquZZrC68s, where RA identifies the module as an RDF graph (independent of its serialization) of the resource, and the remaining characters signify the computed hash of the content.

The Anatomy of a Nanopublication, Groth, 2010, propose to improve the efficiency in finding, connecting, and curating core scientific statements with associated context, with an annotation model and a format based on the RDF language realized with Named Graphs. The Nanopublication Guidelines, 2015, specify how to denote unique RDF graphs for assertions, provenance, and publication information, which make up the body of a nanopublication that can be used as a single publishable and citable entity. Nanopublications can be independently used to expose and disseminate individual quantitative and qualitative structured scientific data, without being accompanied with narrative research articles. For example, hypothesis, claims, and negative results can exist on their own, be identifiable, and reused in different places, as well as embedded in articles.

In Decentralized provenance-aware publishing with nanopublications, Kuhn, 2016, argue that due to publication and archival of scientific results is still based on print-centric workflows and commonly considered to be a responsibility of third-party publishers, there is currently no efficient, reliable, and agreed-upon Web-centric methods for publishing scientific datasets, and therefore a bottom-up process is necessary. To this end, the authors present a decentralised server network with a REST API to store, archive, find, and serve data in the form of nanopublications, where the identifiers for the units of information are based on the trusty URI method. The authors argue that the underlying architecture can serve as a reliable and trustworthy low-level semantic publishing, archiving, and data sharing layer that can be used by different knowledge domains.

Signing HTTP Messages, 2019, describes a way for servers and clients to simultaneously add authentication and message integrity to HTTP messages by using a digital signature. The signing is made by signing the HTTP message body and some HTTP headers at the time of making the HTTP request. As the signature is carried in the HTTP headers, the HTTP message is not altered, and thereby does not require a normalisation algorithm to be used on the message structure.

Linked Data Proofs is a mechanism to ensure authenticity and integrity of Linked Data documents through the use of public/private key cryptography. The digital signature is comprised of information about the i) signature suite that was used to create the signature, ii) parameters required to verify it, and iii) the signature value generated by the signature algorithm. The signature typically accompanies the Linked Data document so that the receiver can verify its authenticity using the available information. In contrast to the Signing HTTP Messages method, Linked Data Proofs requires a normalisation algorithm.

Verifiable Claims Data Model 1.0 is a mechanism to express credentials on the Web that is cryptographically secure, privacy respecting, and machine-verifiable. The core model outlines the concepts for claims, credentials, and presentations that can be used by issuers, holders, verifiers, and verifiable data registries.

The Presentation of Self on a Decentralised Web, Guy, 2017, describes various types of decentralised social systems. As per the client–server model, their arrangement generally falls in the spectrum of tightly or loosely coupled architectures. In these systems, actors control client applications typically through user-agents. The degree of coupling that is exerted on the system varies depending on the class of components involved. More specifically, Why is the Web Loosely Coupled? A Multi-Faceted Metric for Service Design, Pautasso, 2009, proposes discovery, identification, binding, platform, interaction, interface orientation, model, granularity, state, evolution, generated code, conversation as metrics for the degree of coupling in service-oriented systems.

Loose coupling reduces the number of assumptions each party makes about the other when exchanging information. For instance, AWWW includes various protocols for interaction, URIs to identify, and standard data formats to communicate the content. The Linked Data design principles then sets further constraints within this framework by favouring HTTP URIs, and the RDF language. A Linked Data driven system using uniform interaction and information exchange can be viewed to be more tightly coupled in comparison to an arbitrary system that is less specified or unspecified. Corollary, the loosely coupled systems set a higher requirement for using Web specifications emerging through open consensus as well as being widely deployed in order for the ecosystem to function. Otherwise, it would be impossible to communicate if there is no higher-order agreement. The loose coupling of interoperable systems based on open standards however can go beyond the arrangement of clients and servers. For instance, the notion of WebID enables loose coupling of identity, identification, authentication, authorization, and storage.

Decentralised datastore (or dataspaces) under the custodian of different actors implicates organic growth and operation as a by product of autonomous operation. In contrast to third-party controlled centralised systems, decentralised systems can have persistent availability (globally), resistance to external censorship, monopoly disruption, increased interest in privacy and security, and enable free expression with respect to information structure and semantics. In order for decentralised datastores to be effective, both servers and client application have an increased requirement on implementing consensus-based open standards in order to cultivate an interoperable ecosystem.

Shared hosting refers to a kind of web hosting service where the server’s resources are shared among its users. While the hosting has usage limits, it is considered to be economical and have reliability features put in place by the host’s system administrators. Any website with its own domain name can act as an independent node in a decentralised system. However, given that multiple websites reside on the same server, system-wide configuration changes or disruptions on the network usually effects all of the websites. Moreover, there are privacy and security concerns if the information on or transferring through individual dataspaces are not encrypted. Hence, the user places their trust on the hosting service’s administrators.

Personal data service (or store) (PDS), personal online datastore (POD) or pod as a noun, loosely refer to an individual-operated Web space with mechanisms for data storage and management that is capable of providing access-controlled read-write operations for independently built interoperable applications. A generic PDS can be virtually realised with any HTTP server with varying configurations and capabilities offering client-side applications to communicate through open protocols and data formats. Web resources that are only intended for public read entails that the custodian allows resources to be available in response to client requests. Servers would need to factor in an authentication mechanism and authorization rules to determine who to serve the resources to and under which access constraints. A client-server interaction based on a read-write Linked Data architecture has affordances for distributed and semantically interconnected Web resources.

The concepts above generally fall into the spectrum of self-hosting for Web space management. An individual responsible for a PDS entails that they are responsible as well as accountable of its operation, abilities, policies and practice decisions on data, persistence strategies and so forth. PDS makes it possible for actors to safeguard their data by setting authentication and authorization rules for any or class of agents, being able to act as the source of truth for WebIDs and the descriptions in the profiles. Running a PDS fundamentally requires the system to fulfill the registration function for the content it makes available. As archiving is an external function, we expect other services to meet that requirement. The certification function can be fulfilled in a similar manner to registration in that the actors that are involved in the quality-control process can register their own units of information in their PDS in response to target units.

I use the same set of questions raised in Systematizing Decentralization and Privacy to study the design choices for an actor-controlled system based in context of the Architecture of the Web:

How is the system decentralized?

The nodes in the system are the Web resources on servers that make up information such as profiles, articles, annotations and notifications, as well archived snapshots. The advantage of this infrastructure is that once a resource is registered (named and stored), they do not require users for its availability or functionality – a reduced attack surface. Users retain their privacy in that, access and use of resources are encapsulated, and such information about identity or actions taken is not required to spread to the rest of the network. In the case of compound resources including information from distributed resources, like with transclusions, or applications dynamically fetching information from different context, nodes may get to be aware of cross-origin requests. The network topology is by default point-to-point, ie. connections between two endpoints. However, in the case of federated servers, information can be routed and disseminated to the whole network, for example, an actor sends out a notification about their activity. The authority topology is that nodes (client-server or server-server) interact directly with each other, without necessarily requiring third-party nodes to participate. In order to establish trust, a certain amount of transparency can be used to make an authority accountable. In federated designs, users place their trust on their provider nodes who act as authorities on their behalf with other providers (server-server).

What advantages do we get from decentralizing?

There is no single point of failure, has high availability, system performance, and generally survivable. However, the system’s usefulness may degrade if certain resources in the network disappear or become inaccessible. For example, if a notification about an annotation is no longer available, then the annotation may need to be discovered through other routes. There is a flexible trust model in that, there is no natural central authority. Multiple independent authorities exist, and each verify claims based on their own (or shared) decisions eg. self-signed certificates as part of web of trust, or in the case of anyone can say anything about anything, consumers of such information verify the statements based on their needs. There is a distributed allocation of resources assists ease of deployment, and independent extensibility. As with the open-world assumption, nodes are not expected to have complete knowledge, and conflicting knowledge can enter the system. Decentralised systems while benefit from network location diversity, they may also be subject to censored both on the network and legal levels. The development and operation is decoupled in that open software can be publicly visible and auditable before being used by authorities. Original resources as well as versioned or archived copies can have publicly verifiable integrity. Decentralised system are relatively secure in that compromising or weakening the system requires attackers to engage with different services, or code bases. Public accountability can be devised to detect and exclude misinformation, as well as misbehaving or compromised actors. A decentralised system can incorporate trusted centralised nodes eg. community agreed scholarly reviews, archives, caching servers, proxies, and so forth, while if compromised by adversaries or disappearance is temporarily kept local – possibly a performance or discovery problem.

How does decentralization support privacy?

Due to the way resources are distributed in decentralised systems, they offer greater user data privacy with respect to their communication, identities, and actions than centralised systems in that there is no single authority that can observe all data and interactions. In essence, it makes it challenging for an adversary to aggregate and connect all information in a way that can undeniably hold a participant in the system accountable. These systems can allow different degrees of confidential participation in that agents (human or machine) can remain anonymous or use their pseudonymous identities. For instance, sensitive research data can be kept confidential. Access controlled resources can be accessible by trusted parties. Actions of an actor can potentially be unlinkable and so rendering a global search of information useless. The communication between parties can be undetectable and unobservable by unauthorised third-parties.

What are the disadvantages of decentralizing?

Decentralised systems inherently increases the number of attack vectors that an adversary can use due to number of software implementations and configurations. A compromised node in the network can potentially monitor activities, corrupt information exchange, insert malicious code, distribute data to interested parties while undetected, reduce or expose anonymity, provide inconsistent views, and so forth. As there is no single authority to globally manage the nodes in the network, the quality and availability of the services will vary and becomes challenging for each node to actively and effectively protect themselves. For example, an article, its inbox, and the annotations referred from each notification may all reside at different network locations, and thus data consumers would need to factor in potential issues arising from each node by defending themselves or routing around disturbances. Decentralisation also poses an obstacle for accountability and reputation in that while actors may control a node, handling the ramifications of a misbehaved node becomes difficult in the broader ecosystem. Finding, discovering, and interactions on fine-grained resources are typically less efficient than centralised systems.

What implicit centralized assumptions remain?

Decentralised systems include social and technical aspects of centralisations. The DNS is the most prominent example of centralisation in that domain names are administered by an entity other than the actors interacting in the information space. The name which resolves to a node is generally governed by a single authority solely determining the level of privacy the node offers as well as its availability. For instance, since registration of units of communication are mere records of content, their persistence and preservation are at the discretion of the owners of the URI space. Thus, archiving by trusted parties is necessary to take into account – which essentially introduces additional centralised entities. The use of certain certification authorities also introduces hierarchically structured centralised governance into the system, hence fundamentally allowing single entities to revoke authenticate credentials, or eventually influencing abuse-prevention, reputation, and trust. Along these lines, carrying out access policy management and enforcement by independent nodes introduces central locations for potential attacks. Web standards are a form of social centralisation in that they reflect the participants’ norms and agenda. Similarly, content delivery networks, for wide reuse introduces a single point of failure for large number of services – a potential risk even for open-source software.

We can observe different degrees of control; individual and third-party actors at the opposite ends of the control spectrum for identifiers, data, and applications:

Identifiers

Rules and governance of naming things.

Data

Information that can be shaped and used with intended semantics.

Applications

Computer software that performs functions on data on behalf of a user.

On one end of the spectrum, an actor eg. an individual, registers their own domain name, hosts their personal storage, and uses applications of their choosing. With that, they have highest degree of autonomy. On the other end of the spectrum, an individual’s identifier as well as a range of expressible identities, the data location and conditions of use, and assigned applications by third-party services.

I focus on interactions based on the client–server model under the Architecture of the Web, identifiers and data are conceptually decoupled components. In the case of identifiers, I focus mainly on HTTP-based URIs, and the social conventions around URI Ownership. Finally, the applications comprise software that actors can use.

While centralisation of data and applications are two distinct areas of concern, some of the challenges we face in scholarly communication make them appear as one. When we discuss challenges around access to knowledge, the location and quality of the data is often conflated with the tools or services which are required to access the data. Being required to use particular software to create, format or share research articles, to create accounts on particular services, or agree to certain terms and conditions impose a set of characteristics on research output or data itself.

Data and application appear to be intertwined because the workflows that the researchers are required to use tend to be unique and proprietary and predominantly enforced by third-parties. While some of the differences in designs may be due to historical reasons they also ensure various forms of vendor lock-in on both the research outputs and the applications to generate, share and reuse data. The use of non-interoperable (or even post-facto) methods to create and exchange data on the Web leads to the fragmentation of data through silo-services, and along with it the dependency on applications having their interaction mechanisms established via out-of-band knowledge. By strictly focusing on the current quality of machine-readable scholarly information on the Web, the technical obstacles are evident and ample. The applications need to be continuously hard-wired to know where and what to look for; what to do once something of interest is found; or where to go next in the discovery phase. Similarly, automating discovery of fine-grained information without human interference is still central to the limitations.

Different degrees of control would influence the forces and functions in scholarly communication. For instance, while ORCID is community controlled, it is possible for individuals to customise their profiles and to link to their other identities. The profile description can indicate the location of researcher’s personal online storage, inbox, as well as outbox, which can be controlled by the individual. The DOI system on the other hand is governed by an organisation where the mapping between an identifier and the location of content they refer to is managed by select few organisations (as opposed to individuals).

Desired qualities: Based on aforementioned research, standards, and practices in the field towards Web-based personal identifiers and representations, I infer a set of qualities that can support individuals when participating in a decentralised scholarly system:

• Self-hosted and individually-controlled personal identifiers denoted with HTTP URIs
• Profile customisability and connectivity
• Linkable multiple identities
• Identity privacy and control

Standards-based Forces and Functions: While every system of scholarly communication can have its forces and functions arbitrarily configured irrespective of their actual implementation, I have focused on a subset arrangement that is based on existing Web specifications facilitating the manifestation of interoperable information spaces. The model can be extended to include other or future specification with respect to forces and functions in scientific communication.