Identifiers should deliver value through curation

Every institution today mints identifiers on the web, and their web servers resolve them to digital resources. Every institution has an actionable URI space. As the web as a medium becomes increasingly pervasive and reaches into institutional and professional life, user expectations will continue to grow. As more content becomes available digitally, data silos are being opened up for access on the web. As scholarly participation in real-time conversations on the web explodes through blogs, tweets, etc so will the reliance on the HTTP URI as the fundamental currency of value of the web grow.

Steps must be taken to protect the URI as the emergent currency of the web – to organise the URI space with respect to organisational responsibilities and the services offered within it, and to maximise the value generated by those URIs to benefit the institution.

Organisational mechanisms to provide reliability of these HTTP URIs are vital. Automation of curative and administrative responsibilities are beginning to move away from big, specialist systems, towards sets of services that can interact with resources and stakeholders on the web in useful, flexible and granular ways. Repositories are seen less as places or systems, and more as part of a commitment towards preservation and curation.

Table 1 – adaptation of UC3’s model of “micro-services”

The table above shows an adaption of the UC3 philosophy to move towards a services delivery model, one that ultimately links its preservation obligations at the lower tiers to persistently deliver and curate to its obligations to value at the upper ones. The additions are given in bold. UC3 is seeing its delivery less in terms of a monolithic system but more of a process. And these activities rely on identifiers: it is clear that in such a model, a set of agile, granular, and independent services can support changing environmental needs, cost efficiencies can be made, and value delivered if a pervasive underlying “fabric” of identifiers and linkages is carefully designed and managed.

Trust as a value proposition on the web

Organisations try hard to achieve high degree of trust on the web. Commercial organisations such as Freebase and Thompson-Reuters exemplify the desire to achieve as much reuse of identifiers which they curate on the web as possible, and have invested heavily in their identifiers. But these are the areas in which institutions have created a wealth of expertise and understanding, as they have dealt with themes such as origination and provenance, reproduce-ability as the basis of trust and validation, complex data management, origination, processing, and application to real-world artefacts as well as data.

These values of trust and reliability speak to the persistent identifier debate. The embedding and use of the HTTP URI to cater for a variety of hitherto more specialist activities can be seen as a tremendous opportunity for institutions and stakeholders to realise value from what they already own, know and do.

The arguments for ubiquity of HTTP URIs do not detract in any way from the fact that many uses of persistent identifiers demand tightly-specified and tightly-controlled metadata schema, data models or ontologies, thus taking a centralised approach to trust. However, a centralised approach will necessarily be limited in scope unless it addresses interoperability: a motivating factor behind the development of the MPEG-21 RDD (Rights Data Dictionary) was that semantic mappings across metadata schemes ought to be tightly controlled by a central authority. A similar approach has been adopted by the DOI with its use of the indecs Data Dictionary (iDD).

Summary

Solutions and practices to manage the transformational changes required to succeed in this evolutionary web environment are evolving. Innovative technical services based in the cloud have already seen widespread growth. This trend will continue. Institutions are already coming together to cooperate over persistent identifier standards at both national and international levels. JISC is engaged in an ongoing international programme of work to support this, covering persistent repository, author and organisational identifiers[1]. However, the way in which institutions choose to represent what they have and do in this environment remains the preserve of the institution.

There are significant challenges ahead. For institutions, a simplistic view of stand-alone “repository systems” that support persistent identifiers will not suffice.

The informational, technical and social behavioural context in which an institution is situated is dynamic, so the practices and supportive technology must also be dynamic to meet these challenges. The overall focus must remain with maintenance of a commitment towards identifiers as the underpinning of a pervasive institutional presence on the web.

When carefully understood in relation to its core activities, there is good reason to believe the link effect on the web can help institutions to successfully undergo the transformative processes required to let the web deliver value to them – but without persistently reliable, trustworthy and authoritative web identifiers, any web strategy will fail.

“Without adequate institutional commitment and clearly defined roles and responsibilities, identifiers cannot offer any guarantees of persistence, locate-ability, or action-ability in the long or short-term.”[2]

Web technologies and practices have reached a tipping point with respect to persistence of identifiers. So not only do we need to care – we need to care and act now.

This section discusses how developments on the web might impinge on the more preservation-oriented practices around persistent identifiers. It puts forwards a viewpoint that these developments require acknowledgement and action.

Web terms and processes

URLs and HTTP URIs: A Uniform Resource Locator (URL) is an identifier used on the web, and stands for “Uniform Resource Locator”, for instance “http://www.example.com”. The “http:” part specifies the protocol for resolution (others include “ftp:” and “mailto”). An HTTP URI is a URL that uses the HTTP protocol.

Resources on the web and in the real world: A “Resource” is the web term for any entity referred to by a URI. Some resources exist on the web, such as web pages, images, documents etc. Other resources are real-world entities such as people, events, organisations. URIs can be used as identifiers for both of these categories of “things”. If a URI names a thing on the web it is known as an informational resource and the URI resolves to the thing that it names. If the URI names a real-world entity, then it should resolve to a description of the thing that it names, ie metadata about the resource.

Representations and content negotiation: A URI names a resource, but the resource may be available in different formats or representations. For instance a journal article may be available as a web page, a PDF file and a word document. An image may be available as a JPG file and a TIFF file. Content negotiation is the process by which the agent retrieving the resource (typically a web browser) informs the provider of a resource (eg a web server) what kinds of representations it prefers, and by which the provider supplies an appropriate representations.

These processes are important because they underpin the way in which information is made openly available on the web, so that it may be discovered, cited, and potentially interlinked, to form “linked data”. For “linked data” to work, identifiers for non-informational resources need to retrieve descriptions that can be understood by machines. If these descriptions contain URIs for concepts the data is termed “semantic”.

One goal of linking entities this way is to produce a wide variety of useful services that can use data from anywhere on the web in potentially widely different and unforeseen contexts.

There is no central authority responsible for allocating HTTP URIs. It is up to the “owner” of the URIs to take care of them. If the URIs are not appropriately managed this can lead to broken links (“link rot”), where a URI no longer resolves to its representation (or metadata in the case of non-informational resources). Different URIs may be created for the same thing, by different people and/or systems, and unless appropriate processes are put in place it is difficult to have common agreement that the same “thing” is being identified. “Co-reference” is the situation where more than one identifier is used for a single resource.

The web as an institutional problem

There has been a certain persistence of debate about persistent identifiers, and the same themes recur. Studies have suggested digital resources on the web require additional attention to enable their accessibility and reusability over time[1]. Hyperlinks “decay”[2]. There has been a lack of shared agreement over what precise purpose persistent identifiers serve, what they actually should “do”. The scope of Persistent Identifiers might include any or all of naming, resolution, binding to descriptive metadata, current adoption, trustworthiness and organisational hierarchy, human readability and technical interoperability with other identifier standards – it all depends on the underlying requirements that the identifier must help fulfil.

The work and debates around Persistent identifiers has primarily been set in the context of identification of digital resources over time. Implementing support for persistent identifiers has often been regarded as a technical issue that requires a technical solution. A sizable number of divergent systems have been developed to enable Persistent Identification, and a number of schemes developed.

However, this seemingly ever-fragmenting situation may be changing. At the meeting convened by JISC in February 2010 to discuss persistent identifiers, the usual discord as to the “correct” solution was displaced by a general will for agreement amongst parties to commit to areas of commonality, and to move the debate forward, acknowledging the increasing importance of the web as a solution, rather than merely treat the web as a problem to overcome because of link-rot.

The link-rot issue has by no means gone away: but developments and early adopters may be showing the way forward: the web may in fact offer a helping hand, and the responsibilities and organisational commitments institutions need to make “on” and “off” the web may be reaching some clarity.

There has been a shift in awareness that URLs are not just a transitory mechanism for accessing resources, but are fundamentally identifiers in their own right, and need to be managed as such.

Towards the web as an institutional solution

More recently, a number of factors have contributed towards evolution of thinking about Persistent Identifier systems, more towards an increased automated support of persistence of identifiers, in whatever form that may take. For example, The University of California Curation Center (UC3) have evolved a set of “curation microservices”[3] to address the notion that there are many social and organisational functions underpinning their overall remit to keep their managed resources available over time, and it is unrealistic and not cost-effective to expect a single system or even architecture to support this. The intellectual shift from system boundary to responsibility boundary is further reflected in the work of PILIN, which utilised a notion of “curation boundaries” in piloting a national solution for persistent identifiers[4].

These factors are down to evolution of the usage of web itself, as well as a gradual evolution of technical web standards. The web in a sense could be seen to be going back to its roots[5]. There is great power in enabling the “link effect”: keeping to simple standards and affording mass connectivity, coalescing around identifiers, URIs, to identify “resources” on the web, with a common resolution mechanism, HTTP.

It is therefore important to understand how changes in web practices impact the underlying needs addressed by persistent identifiers.

From carriers to content – generalised representation of entities on the web

The rise of social networks and tagging as an integral part of web usage has seen web identifiers used not just for web pages, images, PDF files etc, but to represent real-world people, organisations, places, events and so on. Tags are commonplace in their use for informal classification. Services such as Twitter or Delicious allow URLs for tags to be minted to name anything the user chooses, and to be resolved to presentable information.

There are now sufficient standardisation for representing these “non-informational resources” to enable the “linked data” community to create generalised descriptive metadata on the web, metadata that includes web-addressable identifiers for anything. The metadata ascribing meaning to identifiers on the web may in itself be persistent and on the web.

Linked data, open data, and the semantic web

Despite significant academic research into application of artificial intelligence techniques to the web, with some success as distinct communities are making ontologies available [oceanography] and finding ways of inter-linking them, broader interest has only begun with very recent grass-roots efforts within the “linked data” community. Data is being made available on the web, and practices for their interlinking defined[6].

The exploration of the techniques required to achieve useful results has led to emergent practices revolving around identifiers, especially around HTTP URIs. Debates and solutions range over a number of resultant issues, such as whether HTTP URIs should be “opaque” or human-readable.

These practices include directly supporting, on the web, functionalities that can potentially assist with curation obligations, functionalities traditionally the preserve of specialist institutional systems.

It is arguably worthwhile to pause to consider how web technologies can today cover functionalities traditionally associated with persistent identifiers:

• Discovery through semantic search –SPARQL[7] is a standard capable of querying RDF data over the web. Major search engines are now indexing RDF, and several specialist engines exist such as sig.ma. By indexing data that uses web-compatible classification formats such as SKOS, these services have the potential to enable web-scale discovery more traditionally the domain of specialist thesaurus-based search tools within the institution.
• Text-mining, for automated allocation of URIs to concepts and data found within text [8].
• Representational variants of resources – content negotiation[9] provides a standardised means of returning a representation of some entity in a digital format appropriate to the tool (eg, web browser) requesting it.
• Time-variant versioning of resources – a method from the Memento project[10] for retrieval of different resource versions based on HTTP URIs is seeking standardisation
• Breadth and expressivity of identifier linkage on the web – going well beyond the plain hyperlink, searches using services such as Sindice, OKKAM and RKBExplorer can group co-references, where multiple URIs may denote the same entity. Typed RDF links can categorise the linkages, but also collection-type linkages are possible via OAI-ORE.
• Citation by URN – An international implementation of a URN (URI) resolver exists serving several nations that have adopted the URN-NBN for bibliographic records. The implementation is extremely lightweight but does assist persistent identification for the intended purpose.
• Addressability of software – the widespread adoption of web-addressable “REST-ful”[11] web services using HTTP URIs has overcome a great number of interoperability challenges. The chief point is that today, a great variety of software is available to produce and consume resources and data given its availability via HTTP URIs.

The spectrum of current approaches towards availability and interoperability of data on the web that is intrinsically associated with web identifiers is outlined in this blog post by Paul Wark[12].

Other factors related to the web

Behaviour and social expectations

All stakeholders – whether teachers, learners, researchers, administrators or scholars – today expect to be a citizen on the web, as well as one affiliated to an institution. Being “on the web” is no longer the preserve of web specialists or marketers. Scholarly practices are conducted to some degree or other via the web, and the web can be used to contribute significantly to an institution or individual’s reputation. In some industries, a new class of professional has emerged to maximise value through the use of new services such as Twitter.

Data deluge

For digital items, the main issues are no longer about holding and controlling the availability of a “copy”: copying is cheap As a recent Economist article[13] puts it, the “data deluge” is upon us. “Data about everyone and everything will be everywhere”. In this environment, an institution must establish its reputation through differentiation on the web – there is a need to stand out. Persistent identifiers can help with branding and a clearer ownership of intellectual property through more definitive attribution and citation.

Attention data and Real-time

Linkage on the web is is no longer static. Identifiers within RSS feeds are standard. Usage of web-addressable identifiers embedded within near real-time information streams has become commonplace on services such as Twitter. Indeed, sizeable commercial ventures have turned to this form of identifier to differentiate their web search and better target advertising using data about user’s habits on the web (eg Topsy, a search engine powered by tweets[14]). Timeliness has become increasingly valuable on the web, and offers a new way to engage with scholarly resources, individuals and the institution as a whole.

Impact on the persistent identifiers debate

How far these developments impact the overall remit for persist identifiers remains an open question. Ultimately the answers can only be determined by specific needs, and a determination of the responsibilities amongst the parties involved. The debate over responsibility therefore ranges from the individual researcher, self-publishing their work, through institutional commitments, to national and international levels.

Many important questions remain: for example, there are as yet no clear answers as to the interplay between centralised authority and emergent definitions of metadata and its meanings – ie, should one invest in technically-focussed linked data approaches alone and getting raw data “out there”, or concentrate more on involvement with specialist authoritative bodies, using schemes built using XML, and sacrifice open and decentralised interoperability on the web? To what degree should one reuse or “import” another URI not under one’s domain or namespace, eg a term from a vocabulary, risking conflict? Should one turn to authorities to maintain the appropriate control but lose ownership and flexibility? Who ought to be responsible for disambiguation and co-references present when linking data in unanticipated ways?

Despite these issues, there is a rapid increase in adoption of “Linked Data” practices for data and metadata publishing, practices that utilise HTTP URIs together with W3C semantic web standards for formally representing descriptive metadata on the web. Practices now exist on the web, at some level at least, to support a broad spectrum of needs around persistent identifiers.

A full and up-to-date list of commonly-considered PID schemes is given at http://repinf.pbworks.com/Persistent-identifiers .

Amongst the categories most relevant to the PID meeting discussions were those underlying notions of trustworthiness, robustness and reliability:

• Governance – use of established identifier schemes, conventions, practices and policies
• Reliability –that an identifier can always be used to locate the resource and will persistently refer to the same kind of entity
• Accuracy – the level of rigour applied to description associated with the identifier, formality of its metadata scheme or ontology, whether this is authority-controlled
• Authenticity – knowing when the object referred to has been mdofied, the provenance of both identifier and referent

The more centralised approach was represented at the meeting by the ANDS project[2], whereas a demonstrator from the RIDIR project[3] and Hugh Glaser’s RKBExplorer/CRS[4] approach illustrated areas of progress that can be made directly using decentralized approach to authority, and constructed on top of existing web technologies.

To move forward, the community clearly recognises the importance of adopting different schemes according to particular needs, and that technical implementations are not the main issue. The meeting participants generally agreed that one solution will never fit all requirements, that there will always be multiple technologies to choose from, that choice of identifier scheme frequently depends on choices embedded within a technology.

Exemplary Schemes and Current Approaches

Some technical approaches afford greater automated support for persistent identifiers than others – this is an important distinction when considering use of identifiers within the overall curation context, as it determines the degree to which institutions must invest in methods that deliver the “promise of stewardship” associated with an identifier.

Three examples illustrate the distinctions:

• The DOI – Digital Object Identifier – was established in 1997 and is primarily used in the publishing industry to identify electronic textual resources, although adoption has begun for data resources as well. As well as providing stewardship guarantees through its technical infrastructure, delivered by the Handle system, the DOI is unique in offering a business model offering organisational guarantees. Semantic interoperability of metadata is controlled through a centralised metadata declaration scheme, and there is a similar concept of application profile under central control. Its resolution mechanism is delivered through the Handle system. Technical interoperability with the web is provided via gateways, though this presents challenges with respect to Linked Data practices[5]
• The ARK – Archival Resource Key – is a relatively recent initiative, established in 2001. It is a URI scheme designed natively for the web over HTTP, so is able to utilise standard web machinery for resolution and is lightweight to implement. The scheme is designed to keep separate and explicit within its identifier string structure parts that represent the stable name component, and a stable authority-controlled record of the institution promising stewardship. Notably, reliance on the internet Domain Name System, DNS, is avoided for the purposes of identifying the authority responsible for curation of the object and its identifier.
• “Native Web” – no particular scheme is assumed in this scenario. Any stakeholder can choose to make available and curate resources on the web with no additional scheme or technical machinery; the distinguishing concept here is that no automated support for resource and identifier curation is assumed beyond web standards and generic tools, such as standards-compliant web servers and browsers. As practices evolve that utilise native web machinery, many aspects that underlie traditional notions of “digital object”, such as metadata description, multiple representations and versioning, can be implemented using generic tools. Stakeholders must, however, make their own arrangements for dealing with transfer of name ownership, as all identifiers will use the web’s Domain Name System.

Persistence is not just about identifiers.  Persistence of access to the resource is important – where the resource is stored and what happens if the resource is moved to another location.  Persistence of metadata is important for ensuring that resources can be discovered and described.  Ensuring that these underlying persistence requirements are catered for is part of the Persistent Identifier picture.

Persistent Identifiers have increased in priority because, amongst other reasons:

1. There is a need for Persistent Identifiers for other kinds of entities, beyond scholarly “resources” such as images and textual work.  Persistent Identifiers are needed for a wide variety of curated materials, such as scientific data and learning materials.  Persistent Identifiers are also needed for organisations and people, and the roles they might adopt, such “author”.
2. HTTP URIs are the essential underlying currency on the web.  This was first the case when many HTTP URLs simply resolved to a single HTML representation, allowing construction of effective searches based on page ranks.  Through the Web2.0 phenomenon, HTTP URIs can refer to globally-shared subjects via tags on bookmarking sites, “hash-tagged” events on twitter, and RSS feeds.  With the availability of linked data on the web, the persistence of the identifiers directly determines the validity of this data.
3. As linked data sets are currently being produced, there is a need to establish long-term guarantees around the HTTP URI they mint.  For example, data.gov.uk provides a top-level internet domain name for government data sets.  This top-level domain name will be expected to remain stable over time, along with expectations for all URIs within it also to remain stable.
4. Overall, HTTP URIs need to be “trustworthy”. In an environment where participants are increasingly reliant on digital representations and interaction, the value of a participant may be directly related to trustworthiness of their HTTP URIs.

There are a great many resources available on the subject of Persistent Identifiers and their scope. Some points of reference were created for the Persistent Identifier Meeting [1], and a glossary drafted [2]; recommended introductions are available[3][4].

This page intends to give a flavour of the topic.

What is a persistent identifier?  What is it for?  What are we talking about?

• A Persistent Identifier is typically a string, eg doi:12345/6789…, or http://purl.net/123/example456.
• The Persistent Identifier names some entity, so the organisation “minting” the identifier “doi:12345/6789” is ascribing a name to something, eg a journal.
• A Persistent Identifier should offer a stable “promise of stewardship” over time, of the link between the identifier itself and its relationship with its “referent”, the thing it identifies.  It should not ever refer to any other thing.
• The identifier is “actionable” if, when it is clicked in a browser, the user of the identifier is taken to some “representation” of the thing, such as a PDF representation of a journal article. This process is known as “resolution”.

Persistent identifiers have been around for a long time.

For physical entities, those “off the web”, ISBN[5] is an example of a well-known identifier scheme playing a crucial role in facilitating agreement between parties, typically publishers.  The ISBN supplies a context–free identity to a (class of) book.  To order a physical copy of a book, one can order it using the ISBN.  ISBNs are commonly associated with metadata, and permit cataloguing using well-known rules bound by organisational agreements.   An authority looks after the overall allocation and quality of ISBNs.  The ISBN is one of a family of standard identifiers that include ISSNs[6], that are reliable enough to enable citation.

JISC funded the RIDIR project with the intention of understanding identifier interoperability, particularly what could be learned from efforts undertaken by the ISO for established schemes such as ISBN and ISSN and interoperability with other identifier schemes[7][8].  The overall outcome has been to take a broader look at the subject of persistent identifiers, particularly within the context of emergent practices around the HTTP URI as the basis for linked data on the web.

The ISBN illustrates that an identifier is some string, an information-bearing object that should guarantee its users a ‘promise of stewardship’ for the thing it identifies.  It should also allow its users access to metadata describing the thing referred to.

For curated digital resources such as images or texts stored electronically, an actionable identifier ought not only to uniquely “name” the identified object, but permit retrieval (resolve to) some useful representation of it.  As this digital information on the network can be easily altered or replaced, the bundling of descriptive metadata with the information itself has become a standardised practice within the archival community to support persistence, and is referred to as a “digital object”.

Historically, the debate has been one of disputes as to the “correct” solution to the problem of persistent identifiers . However, a general theme emerged during the meeting that this debate has matured significantly, and that it would be more productive to pursue areas of commonality and agreement, whilst recognising the heterogeneity of the landscape. Reflecting the growing importance of open data on the web, five candidate statements of common agreement were proposed as a basis for moving forward; in summary they recognise the heterogeneity of the environment, encourage the use of dominant schemes where appropriate to a particular context, and otherwise encourage the ubiquity of HTTP URIs, provide agreed and consistent HTTP URI manifestations for schemes that do not have them, and encourage higher levels of good practice for HTTP URI schemes.

Landscape

The meeting participants gave a number of use cases covering a broad spectrum of needs and areas of value that would be met by improving practices around existing persistent identifier schemes. Presentations reflected aspects arising from differing approaches and implementations. For example, addressing requirements using recently clarified web architecture standards has ramifications with regards to trust, ownership and practices needed to manage web HTTP URIs; the DOI remains popular in certain contexts, and an organisation using DOIs will need to commit to practices arising from a fee-supported organisational model, as well as to centralised technical identifier standards beyond those pertaining to the web. The RIDIR project demonstrated how useful services for repository purposes can be built assuming no particular centralised identifier infrastructure, and tackling aspects around the meanings underlying “identifier interoperability”; the Australian National Data Service (ANDS) illustrates how the Handle system can be used to implement common identifier services, at national level, to tackle certain HTTP URI issues; SURF and Knowledge Exchange emphasised the organisational and the global perspective over the technical ones, discussed trust and a specific implementation of a pragmatic resolver for URN-NBNs; Southampton have developed approaches to deal with multiple identifiers and co-reference, in a linked data technical setting that emphasises decentralised ownership of HTTP URIs.

Issues and Priorities

The chair, Chris Awre, gave an overview of the issues around persistent identifiers from an institutional perspective. Participants recognised that what constitutes a “good identifier” depends on the underlying needs, and that these varied. Factors included the type of entity; varying curation responsibilities to preserve resources and access to them, together with their descriptive metadata and networks of relationships; scalability in terms of volume of entities, granularity and structure of the entities that need identifying; those factors contributing to trust, and various issues surrounding privacy and rights.

Priorities varied across groups represented, covering Research Papers, Research Data, Learning Materials, Cultural Heritage and Administrative Information. Some examples are that: self-published materials and grey literature could be seen as reliable, reputation of creator enhanced, and impacts can be reliably measured to provide business intelligence and agility. Authors want to trust the persistence of the identifier and understand the persistence criteria; they want the service obligations explicit. Some deem the HTTP URI important in terms of branding. Scientists and publishers want to reliably curate references to their datasets, which can be complex and require persistent identifiers in different ways to address granularity and structure. Teachers do not tend to share learning materials currently, possibly because they have no clear incentive to manage and make available materials themselves, there is a lack of metadata available. Institutional administration would benefit from being able to reliably track resources, identify modules taught and build up historical data to generate credibility and avoid the current loss of corporate memory. PIDs would help audits and help support unify approaches to regulatory compliance; better planning can be done across the institution. Beyond identifying resources, a reliable and interoperable means for identifying people and roles, such as author and contributor, and staff and student, is required. For some, support for persistent identifiers amounted to bringing order from chaos.

Ways forward

At a technical level, JISC should help create facilities rather than authorities, decentralised services that use independently-managed identifiers, rather than centralised services that create and manage them. Examples of concrete services where JISC might usefully play a role include broken link services and co-reference services.

Evidence needs gathering of where persistent identifiers work well in practice. Participants indicated that case studies and demonstrators that situate technical solutions in context, articulate specific workflows and promote good practice in managing identifiers would be of considerable value[1]. Toolkits and frameworks would be useful to institutions to address issues of cost and perceived cost of the choices available. Awareness needs to be raised of existing achievements, the options already available and how to use them effectively. The heterogeneity of the factors involved in persistence of identification needs to be considered, particularly in the context of variations between stakeholder needs. Wide dissemination of materials produced by this exercise is needed.

Suggestions for where JISC may usefully play a role included working with bodies such as UCISA, SCONUL and the Research Councils to develop and provide advice on policy and illustrate the appropriate business cases and uses for persistent identifiers.

Using existing channels where possible, JISC could assist in raising awareness by engaging with stakeholders, both within the HEI sector and with external organisations. By doing so, and noting that the issues are not merely technical but largely organisational, JISC could start to impact on practice.

JISC should identify those for whom persistent identifiers really matter and engage with them in a role of facilitation rather than intervention. This could lead to a common definition of what constitutes a “good identifier”, internationally and cross-sector.

Presented below are the main (unattributed) tweets captured during and shortly after the meeting, in chronological order.

CrossRef was an enabling factor for DOI

DOI is not perfect, but it won’t die

Desire for more than a re-hash of Cool URIs

DOI is an implementation ofan HTTP redirect mechanism, that is its technical value

DOI has social values – these could be added to aother HTTP redirect mechanism

No URI is cool on its own – people have to choose to make and keep URIs cool

Redirects are “rubbish” for browser bookmarks (the persistent identifer is not there for bookmarking)

Why not just Google for it (do we need persistent identifiers?)

PURLs are good but DOIs have associated metadata, which is better

DOIs may not be better from a techinical point of view, but they have social traction – this is important

DOI success may be based on marketing to people who didn’t understand the real issue

Understanding why DOIs are successful will help in the design of persistent identifiers iin the future

How much does DOI metadata get used, and by whom?  It is used by ePrints reposotiries when importing new items.

DOIs are successful because they are maintainted by a community to whom the identifiers are business-critical

Some publishers do not understand that the DOI prefix is not brand-related and mistakenly change prefixes on takeover, CrossRef works to avoid this

What is OpenURL – an identifier or an encoded search?

The only persistence of authority in an identifier lies in organisations and services providing it (Weibel)

DOIs are not HTTP URIs, but http://dx.doi.org gives us URIs

URIs are names, not addresses

HTTP URIs are vulnerable – DNS for resolution, web servers for storage and delivery

Opaque identifiers are demotivating.  Branding is important

FRBR is a good place to start for ontology

FRBR is the wrong place to start as it is an ontology of the publishing workd and not those doing scholarly communication

If a problem never goes away, maybe it’s not a problem.  No solution, just compromises

Don’t make promises you cn’t keep – new naming schemes without credible service guarantees is irresponsible

Make it easy for ordinary users to mint good URIs themselves.  make it easy to author and serve metadata.  Think about indexing and replication

Use bottom-up identifier schemes, rather than top-down.  That’s the way linked data works

Motivate people to share

Fix the hole where the rain gets in – persistence of domain names

Peer recognition counts – use social networking tools

FRBR is relevant on what to identify, especially appropriate copy issues

FRBR has lots of baggage and lots of holes;it attempts to draw too tight a scheme around each level resulting in duplication and holes

Is there anything better than FRBR as a starting point for an ontoogy of works

DOI is a business proposition

DOI could be registered as a URN (later note:  DOI is registered in the info: URI scheme)

Can DOIs work for data – scalability issues, size, number, rate of minting, rate of change

DOIs do have a divergence from web architecture, but it does soerve some communities well

Big advantage of DOI, ARK and other arms-length schemes is they persist when the domain name is lost

If a resource identified by a Handle moves, don’t mint a new one, update the old one – indication of a failure of policy when this happens, and no scheme can survive this

Why have an indirection layer if you aren’t going to use it?

All technical schemes suffer from the same fundamental flaw – lack of responsibility.  No technological solution

NASA areospace documents are a successful example of persistent identification

Search can function as a resolver

Look at what organisational structure each approach needs to work – some solutions may technically be best but be at odds with the reality of organisational requirements

Handles for “made public” data, DOIs for “published2 data?  The difference is governance and hosting.  Consumer perception is the only difference.

How much does consumer perception (confidence?) matter?

Andrew Treloar:  Identifier, Resolution, Biding, Object

ANDS has no policy yet for PIDS for changing objects

Is it really a problem to have separate PIDs for copies of objects in separate repositories

Authority is established, not bestowed?

Separate PIDS is an issue if you are trying to count your citations

Google scholar might show a solution to separate PIDs for the same work

Is the use of a DOI a public statement of lack of confidence in the future of your organisation?

Use of DOIs is more of a statement of lack of confidenc in the organisation chart as reflected in the domain URLs

Use of DOIs is a recognition that change happens rather than a reflection in the future of the organisatino

DOIs is more of a recogntion of the reality of trading with commercial assets

When journal X moves to publisher Y, how do Cool URIs help, how does DNS help?

Jornal X moving to publisher Y – old publisher redirects URIs to new publisher – the same as telling CrossRef that it has moved

DOI works within a trading system, it’s not necessarily right for other domains – eg research data

Coexistence of DOI with Linked Data would be good

Content negotiation has been around for a while, but it is now very much back on the agenda

People do not understand DOIS, but they do understand URLs

Spirit of compromise and pragmatism in the meeting is good – felt like progress

TheW3C Technical Architecture Group (TAG) identifies two top-level types of entities that need identifiers on the web – informational resources, such as web pages, documents, and data accessible on the web; and non-informational resources, real-world entities such as people and organisations

Resolving identifiers

A fundamental difference between identifiers for informational and non-informational resources is what they resolve to.  Informational resource identifiers by design resolve to a representation of the resource

Persistence

If a persistent identifier is to succeed, the entity to which it refers must itself persist as a general requirement.

Responsibilities for persistence of entities

Alongside the need for persistence of entities, the responsibilities associated with preserving and curating these entities is fundamental – without these being clearly identified, preservation is less likely to be persistent.

Responsibilities for persistence of resolution

Although ownership of entities may change together with their responsibilities for persistence, access to resources can be maintained through a suitable choice of persistent identifier scheme designed to cope with this situation.  Maintaining this persistence of access, by associating an identifier with an entity’s location, brings out a new responsibility.

Hugh’s presentation was based on the notion of distributed authority – what to do when authorities collided, what is the granularity of an authority, what to do when we have PIDS and how to manage the infrastructure.  Hugh established that each metadata owner publishes its own identifiers, and that is the right thing to do as their workflow depends on it, they own the concept of the entity, and there would be problems dealing with change if they used someone else’s.  Hugh gave a demonstration of the Co-Reference Service (CRS) and RKBExplorer which aims to deal with the challenge of many identifiers with metadata published by many authorities by comparing metadata across services to establish a notion of identity and equivalence, and to establish relationships between entities through a linked data approach (though not confined to linked data).  Hugh suggested that JISC could facilitate CRS deployment, take co-reference discovery initiatives and make their data available as linked data.

Bas presented an overview of the persistent identifier activities for SURF and Knowledge Exchange.  The SURFShare programme(2007-2010) is based on the Digital Academic Repositories Project (DAREnet) in the areas of interoperability, communication, registration, sustainability and dynamic archiving.  The programme is aimed at apparent changes in research and scholarly process in the areas of workflow and collaboration, and enhanced publication, addressing an infrastructural requirement for an integrated cooperative research process (seamless data exchanges).  Bas gave an overview of the international nature of the work and the persistent identifier activities that the programme has worked with.  The conclusions and outcomes are URN-NBN is widely used by national libraries and may have potential advantages, that talking will not progress the development of a global URN-NBN based global infrastructure, and that further analysis of future developments is needed.

Bas outlined the Knowledge Exchange (200-2009) programme, based on interoperability of digital repositories (complementing DRIVER II).  The main outcomes are that KE should commission a study to increase awareness of PID important, address the current state of affairs, roadmap and requirements for global resolving.

Bas presented the Global Resolver Service Infrastructure, based on the beliefs that there is a need for a global persistent identifier system, that there will not be one worldwide dominant persistent identifier system, that technology is not the most important challenge – agreed policies and governance are, that action on an international level is needed, and that the URN-NBN system needs to mature rapidly.  The use of URN-NBN is based on it being in use in a number of collaborating national libraries, it is based on trust, it is non-commercial and has an open, inclusive approach, aiming to resolve other systems’ PIDS as well.

Andrew’s presentation was on the ANDS “Identify My Data” service and related issues.  He explained that the function of ANDS was to establish the Australian Research Data Commons, with a vision of more researchers re-using more data more often.  IT builds on the work of PILIN through the guides produced, the pilot infrastructure, the software toolkits and the sustainability report (which did not lead to a national service).  Andrew  explained the motivations for making data public and the benefits of publishing data.  He went on to explain the main verbs that ANDS uses for sharing data.  He explained how using a Handle approach gets around some of the common issues around using HTTP URIs and around organisational and platform issues.  Four kinds of persistence – identifier, resolution, binding/mapping and object – were identified.  A demonstration through screenshots of the ANDS application was presented, and future directions of ANDS were stated.

Steve gave an overview of the RIDIR project, identifying that the main objectives were to engage with the identifier and repositories communities to understand their requirements and to build a fully working demonstrator, and to raise awareness of persistent identifier interoperability issues.  He stated that the project was not about shared persistent identifier services themselves as such services would be available, but was focussed on the use of identifiers.  The RIDIR project found that there was little commonality in requirements, that it was not a priority and the project was exploring areas that were as yet not perceived as concerns.  A demonstration of the “Lost Resource Finder” service was presented, which allows a user to be redirected to the new, correct location for a resource when an identifier is broken.  The new location can either be specified by a repository manager, or it can be based on “crowd intelligence” – based on other users of the service searching for and discovering the new location of the resource within the service.

Paul examined the Handle / DOI approach, stating that DOI is a business proposition.  He gave an overview of what DOIs are, the syntax, metadata and services around DOIs.  The business model was examined, and the current coverage of DOIs was presented in terms of the entities they are currently minted for.  He examined the general requirements in the UKHE sector and how DOIs might represent a value proposition, concluding that DOI cannot and should not be ignored, that it satisfies the concerns of particular domains, and that the growth of DOIs will mean that there will be a ‘mixed economy’ for some time to come.

Henry examined the HTTP URI approach for persistent identifiers, presenting some basic assumptions on the sharing of resources and what URIs are for and exploring what it is to name something.  He presented definitions of desirable characteristics in persistent identification.  He identified that FRBR is a useful starting point for thoughtful ontology in the issue, and concluded with some concrete recommendations on minting good URIs, making it easy for people to author and serve metadata, the gathering and indexing of metadata and representations, motivational factors.

Chris’ presentation can be found at: http://docs.google.com/leaf?id=0B93tJ2TZe3khYWQxNmQ0ZWYtNzcwZi00ZDZhLWIwNjMtNzc3YTk4Y2IzYzQ0&hl=en

Chris’s presentation provided an introduction and an overall context for the meeting, examining why the meeting was taking place and the background initiatives that had led to the meeting, specifically the work done by the RIDIR project and the growing interest JISC has in persistent identifiers.

Amongst other things, Chris identified that

• Identifiers are not uppermost in the consciousness of users of digital content
• Content is becoming lost and hidden.
• There is Increased awareness of identifier schemes, and that something can be done about the issues
• The range of options available can be bewildering, and there is a reliance on identifiers promoted through the technology used
• A focus on identifiers per se can be distracting: purpose, role and persistence itself is important
• Identifiers do not stand alone, their associated meaning and relationships are important.

Persistence itself was highlighted, particularly organisational persistence and the need to record changes, relationships and provenance.

Some suggestions were made on the areas in which JISC could be involved in making persistent identifiers work.

A cultural change programme with the Universities and Colleges Information Systems Association (UCISA) and the Society of College, National and University Libraries (SCONUL) etc could be initiated to encourage people in HEIs to care about identifiers.

JISC should develop a “group of common statements in which we find common agreement” (see Section 4 above, Candidate Statements of Common Agreement).  These can then be used to initiate conversations with the W3C (Technical Architecture Group, RDF) the International DOI Foundation and other parties, and  UKOLN could take a major role in this.

It was stated that there ought to be a common definition of a “good identifier”, internationally and cross-sector.  Engagement with the relevant parties could lead to the development of such a common definition.

The Research Papers group raised an issue of engagement in terms of how to involve researchers who did not want responsibility.

In general, the groups identified the importance of the relationship between incentive for maintaining identifiers and the responsibility for doing so, and closing the gap between them.  Recognition and reward are part of the scholarly process (raised by Research Data) and there is a relationship between online reputation and adoption of identifier schemes (raised by Research Papers).

Getting data published and being respected was an incentive identified by the Research Data group, which emphasised needs for citation and re-use of data in the sector.

The Research Papers group considered personal incentives for maintaining identifiers in more detail, noting that non-opaque identifier schemes help maintain a brand, which therefore suggests use of HTTP URIs as persistent identifiers may be of considerable value here.

The Learning Materials group suggested that the Quality Assurance Agency for Higher Education (QAA) could look beyond the processes involved in teaching and give credit to high quality learning materials as part of their evaluations of HEIs.  This could lead to an increase in sharing these materials, with the lack of sharing being identified as a barrier to adoption of persistent identifiers.  Similarly the HE Academy Professional Recognition Scheme could use the route to career progression as an incentive for identification and sharing of learning materials.  Identification of individuals through persistent identifiers was noted as a potential requirement.

It was stated that the community lacks widely understood policies, and that there are cases where there is a lack of awareness of how to best use existing identifier schemes.

A particular case highlighted was a situation where Handle is in use.  When a resource’s location changes, rather than updating the existing Handle for the resource a new Handle was being minted.

Potentially this indicates the need for a better awareness within HEIs at a policy level (perhaps of existing materials on the use of Handle in this example), so that management of identifier infrastructure can be better implemented at a technical level.  HEI IT Infrastructure’s role is one of exposing information, and strategies need to be in place to support this.

The lack of awareness could also be addressed by specific case studies illustrating best practice usage of existing identifier schemes (see above).

It was noted that persistent identification issues are not merely technical but also organisational, and that by raising awareness and engaging with the community JISC can start to impact on practice.

Where possible, existing channels such as the Repository Support Project (RSP) and the Digital Curation Centre (DCC), should be used.

The Digital Economy Bill may have implications for persistent identifiers in relation to linking to publicly-available material. As the bill goes through the parliamentary process the implications or otherwise of this will become clearer.

Metadata describing the context in which an identified entity is used was identified as an underlying requirement for Learning materials, to record which courses the materials were used on for example, and for Administrative information , to record historical information about modules taught and programmes of study.

Learning Materials also identified that poor modelling and no clear lifecycle were factors that impact the availability of good quality metadata.

The Cultural Heritage group identified various needs that rely on metadata, in areas around access, aggregation and reuse.

Relationships between “versions” of entities was identified as a requirement by Research Data, particularly in the context of dealing with updates and dynamic data, and  Research Papers identified that versioning also represents a problem faced by their sector.

Aggregation of entities into identifiable composites was identified as use case by the Cultural Heritage group.  Learning Materials identified a similar use case in dealing with complex, composite learning materials that can be multi-level, with no linear ordering.

The Administrative Information group identified the need to support multiple views on the same data, recognising that some views may incorporate aggregated views of data present in others .

The Research Papers group identified a need to deal with resources where the responsibility for curation largely lies outside the control of this group, particularly grey literature and self-published literature.  These currently have little by way of curation policy, where publication rather than preservation is seen as the primary objective, despite often being heavily used.

The group identified that responsibility for management of entities can change to reflect changes around people and organisations, and that HTTP URIs may therefore not necessarily be a robust approach for this group.

The Research Data group identified a preservation need: publishers are insisting that data is made available, and there is no clear identification of responsibilities for the preservation itself, this being left up to the researcher/institution who may not yet have defined a preservation strategy.

Responsibilities for maintaining authenticity were also highlighted by this group – being able to ensure that data, once published, has not been tampered with.  Re-use of data is a driver for this group and persistence of access to data through its identifiers also needs maintaining  to support this.

The entities of interest to the Learning Materials group have a variety of associated preservation and access responsibilities, as materials are stored in VLEs such as Moodle^[1] and various repositories, and materials are distributed over the web using services such as Slideshare^[2].  In general, this sector’s entities tend to be “made public” rather than “published”, which suggests a lower importance is attached to preservation responsibility.

Sustainability was identified as a requirement by the Learning Materials group – teachers require learning materials to be available the next time they teach a course, and learners require them to be available when they wish to revise a topic.  As learning materials tend to be distributed over disparate systems, a reliance on these systems for maintaining persistence of access means there is potential for identifiers to break.

The Administrative Information group identified a wide range of metadata to support its needs which is typically held in a range of diverse systems at differing levels within the organisation.

[1] http://moodle.org/

[2] http://www.slideshare.net/

The Research Papers, Learning Materials and Administrative Information groups all identified a need for real-world entity identifiers, particularly people identifiers. The Learning Materials group identified a range of other related real-world entities such as modules, courses and lessons whilst the Administrative Information group has a need to identify records, people and organisations.

Determining “what makes a good identifier” was expressed as a desirable goal at the meeting.  The  headings below represent approximate groupings of statements into what are arguably chief factors  in determining a “good identifier”:

Clear, easy-to-understand advice should be provided on how an HEI might choose between identifier schemes based on their own needs and contexts.

The pros and cons of various approaches in different circumstances, for different purposes, should be outlined and advice should be provided on the adoption and management of the various identifier schemes available.

• Illustrate the benefits in “closing the gap” between those with an incentive  to maintain identifiers and those whose responsibility it is to do so

• Illustrate costs and benefits by articulating workflows in specific HEIs, and identifying stakeholders and value in each context

• Provide a cost assessment of getting a system running using various persistent identifier schemes and infrastructures

• Illustrate how existing software packages that presuppose particular schemes and policies with regard to persistent identifiers work in practice, identifying the implications of the schemes and policies used and addressing any issues arising out of these

• Illustrate the pros and cons of specific approaches and identifier schemes and how well these work across different contexts and communities

• Illustrate best practice usage of existing persistent identifier schemes in various contexts, and where relevant, pay heed to those areas thought to be a source of confusion currently or about which there is a lack of awareness.

• Specific studies from breakout group discussions:
  • The Administrative Information group suggested Southampton University’s approach to identifying people (amongst other things) be the subject of a study.  This uses a system backed by “linked data”, and uses a co-reference approach to relate a multiplicity of identifiers that represent the same person..
  • A case study based on the Research Papers sector may be useful, as this sector deals with a wide variety of curation and preservation responsibilities, many of which lie outside the institution.

HTTP URIs can be and are being used successfully as persistent identifiers, and  Southampton University’s use of Eprints for bibliographic data was raised as an example. In Southampton’s case HTTP URIs (URLs) are redirected to look up syllabus and year for example; change management works well and this is an example where URL policy can work throughout parts of the organisation.   W3C technical reports provide another example.  In this case HTTP URIs embed text strings to represent dates, which takes effort and there is some resistance to maintaining it, but it is a system that has worked well for more than ten years to support discovery of and access to the reports.

It was suggested that JISC should have a role in gathering evidence of where persistent identifiers worked in practice across a range of different contexts to identify, disseminate and support good practices for these situations.  Success stories that illustrate and underpin guidelines in different contexts should be developed and case studies produced that address the needs and barriers expressed during the meeting.

In general JISC should play a role in enabling HEIs to provide services to researchers that they want, using identifiers.  This could perhaps be in the area of preservation services, although the actual implementation may well be at the institutional level, facilitated by JISC’s provision of advice, toolkits and frameworks.

The importance of curation was identified, particularly in terms of addressing a need to provide easy ways for authors to engage with curation practices when self-publishing.

There was a request for “appropriate copy” to be considered an explicit HEI requirement, solutions to which might be implemented as a service based on identifiers.  Again this need may be best addressed by the provision of advice, toolkits and frameworks rather than services provided by JISC.

Both the RIDIR project’s “broken link resolver”, which helps users resolve broken links based on both owner-managed and user-generated input, and the RKBExplorer developments at Southampton, which deals with ambiguity and co-reference using existing URIs, were identified as example services.  The PILIN project also examined the “broken link” issue and identified a Persistent Citation Resolver service using reverse lookup on an identifier that no longer resolved.

It was also suggested that JISC should add further value by working with established services that already use identifiers, the example discussed being PIRUS2 and the JISC Usage Statistics Portal.

Two of the breakout groups expressed needs around services that use identifiers:

• One objective expressed by the Learning Materials group was to enable a “digital repository environment” for learning materials that needs to recognise that materials tend to be distributed widely over the web and not just within repositories.
• The Research Data group stated that publishers are demanding that data (identified in a publication) be made available persistently.  However, the publishers are not taking responsibility for the preservation and identification of the data, the responsibility being left to the authors.

However it was also noted that JISC might play some role in the establishment of a national shared persistent identifier infrastructure service; this might be for example, a “service of last resort” for cases where” all else fails”, whether due to policy or technical reasons.  It was noted that the PILIN project had identified this service as being one primary function of a shared national persistent identifier service.

• providing services around persistent identifiers;
• providing advice, toolkits and frameworks; and
• engagement with the community and increasing awareness.

1. We accept that we are dealing with heterogeneous identifier environments.
2. Well-defined contexts may have their own schemes.  Where dominant schemes are used in a particular context (eg DOI) then we should encourage their ubiquity.
3. Where there are not dominant schemes, then we would encourage the adoption of HTTP URIs
4. We want to ensure that actionable HTTP URI manifestations are available of any non-native HTTP URI identifiers.  We want agreed and consistent ways of constructing these manifestations. We would expect to be able to build services on the basis of these manifestations.
5. Higher levels of good practice may be possible for native HTTP URIs.

These three recurrent high-level themes which represent the overall context of persistence can be summarised are as follows:

• Preservation: The physical location and storage of a resource, responsibilities for its curation, and whether these responsibilities are themselves persistent or likely to change over time
• Metadata and relationships: The ability to describe a resource by associating metadata with a resource’s identifier, to identify the type or kind of resource and to draw relationships between resources by using their identifiers
• Access:  The relationship between a resource’s identifier and the location of that resource; and the services and systems necessary to support this so that the consumer of an identifier can be guaranteed persistent access to the resource.

• to validate the essential points and the common themes distilled from the meeting
• to solicit feedback from meeting participants, and to invite additional comments not elicited in the meeting

The overall background and context for the meeting is given in the Briefing Paper and in Chris Awre’s introductory presentation, with the primary objectives being:

• to identify common features of the landscape where JISC investment might yield benefits; and
• to identify what may be needed or required for the UK landscape

Taking an institutional perspective, Chris examined what we mean by the overall context of persistence in identification, and noted amongst other things that:

• Identifiers are not uppermost in the consciousness of users of digital content
• Content is becoming lost and hidden
• There is Increased awareness of identifier schemes, and that something can be done about the issues
• The range of options available can be bewildering, and there is a reliance on identifiers promoted through the technology used
• A focus on identifiers per se can be distracting: purpose, role and persistence itself is important
• Identifiers do not stand alone; their associated meaning and relationships are important.

Persistence itself was highlighted, particularly organisational persistence and the need to record changes, relationships and provenance.  This is in the context of well-established practice in some domains, such as the use of DOIs, and considerable momentum elsewhere (e.g. from data.gov.uk) behind linked data and HTTP URIs.

Martin Dow
Stephen Bayliss
Acuity Unlimited
Tel: 020 7100 5625
martin.dow@acuityunlimited.co.uk
stephen.bayliss@acuityunlimited.co.uk