In the 21^st century, science is increasingly assisted by computers and other digital machines like remote sensors, digitization pipelines and genome sequencers. In this digital science persistent identifiers (PIDs) play a role to identify people (researchers, collectors, technicians), their organizations and the things they work with (their research inputs and outputs - data, software, literature - and artefacts such as instruments and other physical objects). In an easy-to-read introduction, Meadows et al. (2019) explain the role of PIDs and their value as foundational elements in the overall research information infrastructure. Critically, PIDs act not only as identifiers but also as connectors – of one identified thing to another where a relation can optionally be expressed – ‘A is-employed-by B’, ‘C was-published-by A’, ‘D was-sequenced-by E’, etc. Such connections can be followed both by machines and by humans, not only revealing and providing access to a wide corpus of associated information, but also allowing relations to be understood, insights to be gained and conclusions to be reached. Importantly, the mechanism helps with attribution and recognition of personal and organisational scientific achievements. PIDs make it easier to evaluate the impact of publicly funded work. As noted by Meadows et al. (2019), “DOIs for publications are a great example of how this works in practice”. But they also note that extending the use and support of persistent identifiers further requires stronger community commitments.

One such commitment that leads to de facto wider use of PIDs is commitment to the FAIR Guiding Principles (Wilkinson et al. 2016, Mons et al. 2017) on making data Findable, Accessible, Interoperable, and Reusable (FAIR). Principle F1 states that both data and metadata should be assigned a globally unique and eternally persistent identifier, while principle A1 states that data and metadata should be retrievable via their identifier. Adopted by the European Commission (EC) as a pillar of their policy objectives to advance the global open science movement, FAIR has become one of the strategic priorities for developing the European Open Science Cloud (EOSC) as well as an essential mandate for all European research infrastructures (European Commission 2018a, European Commission 2018b, European Commission 2019a, European Commission 2019b, European Commission 2019c, Schouppe and Burgelman 2018). Meeting this EC mandate (i.e., exhibiting ‘FAIRness’ as a characteristic) occurs when a research infrastructure such as DiSSCo achieves and maintains ‘toto genere' *2compliance with all fifteen of the FAIR Guiding Principles.

As a strategic priority, persistent identifiers are at the core of the EOSC interoperability framework (European Commission 2021b) with a policy governing their use (European Commission 2020). Also mentioned is the “the atomic entity for a FAIR ecosystem”, the FAIR Digital Object (European Commission 2018b). These FAIR Digital Objects, as explained in detail by De Smedt et al. (2020) are self-contained, typed, machine-actionable data packages unambiguously identified with persistent identifiers.

By choosing Digital Object Architecture (Hardisty et al. 2020) DiSSCo achieves FAIRness by default and aligns to the main European science-research infrastructure policies that recognise the value and impact of both persistent identifiers and FAIR Digital Objects i.e., Report on turning FAIR into reality (European Commission 2018b), the EOSC implementation plan (European Commission 2019a), and the EOSC interoperability framework (European Commission 2021b). Everything is identified; thus, everything is findable. Everything is described, and these descriptions (metadata), including how to access things are themselves identified. Persons (users), organisations, workflows and machines are identified. Thus, access can be controlled based on identity. Specific typed definitions of ‘FAIR Digital Object’ (again, each uniquely identified to avoid ambiguity) such as Digital Specimens, Digital Collections*3, Loans and Visits Transactions, Annotations constrain the heterogeneity and incompatibility of data types across systems, thus contributing to interoperability and ensuring the possibility of bidirectional and interdisciplinary interactions between communities such as our natural sciences community and others. By leveraging the first principles and simplifying complexity using persistent identifiers and definitions of FAIR Digital Object types, the resulting objects are available for reuse (Lannom et al. 2020).

A Digital Specimen is a specific kind of FAIR Digital Object that acts on the Internet as a digital surrogate for a physical specimen in a collection. A Digital Specimen, referenced by its unique persistent identifier represents the sum of digital information on the Internet about a physical specimen object in a natural sciences collection and other data derived from or associated with that specimen. A PID for a Digital Specimen complements identifiers of the physical specimens themselves and/or their corresponding digital database records in institutional collection management systems. Such identifiers include: CETAF Stable Identifiers (Güntsch et al. 2017), International Geo Sample Numbers (IGSN) (Lehnert et al. 2019), GUIDs, Darwin Core Triplets, institution/collection codes and catalog numbers.

As a new kind of object alongside natural objects and fabricated tools (Hui 2012), machine-actionable Digital Specimen objects on the Internet are amenable to processing by and transport between heterogeneous information systems. Such digital objects are editable, interactive, reprogrammable, and distributable (Kallinikos et al. 2013). Interoperability difficulties are much reduced by the definition mechanisms for object types and operations that underlie the concept. Such objects have the implicit capability to remain findable, accessible, interoperable and reusable (FAIR) over timescales familiar to collection-holding institutions. This is many decades (100+ years). As a kind of FAIR Digital Object, Digital Specimens assist to integrate collection data in the data rich world of (inter alia) the Earth System Sciences and Life Sciences.

Persistent identifiers that unambiguously identify Digital Specimens (Fig. 1) are one mechanism for digitally transforming collections-based science*4. They are are integral elements of the modern-day open science value chain. Through their sustained persistence PIDs contribute to building and maintaining long-term community trust in the accuracy and authenticity of the scientific data being managed and presented. Not only are they valid over the very long timescales common in the heritage sector but they are also capable of transcending changes in underlying technologies of their implementation. They are foreseen as one of the mechanisms for widening access to natural science collections, for transforming collections-based science into the digital era and for helping this community to fully embark on open science practices.

Figure 1.  

Digitally transforming collections science with Digital Specimens and persistent identifiers (PID).

Persistent identifiers make it possible to reliably refer to and find the digital equivalent of a specific specimen held in the collection of a specific institution, and to reliably access data derived from that specimen. This is especially important in the case of voucher specimens, which are the representative samples of identified organisms providing the verifiable evidence for authenticating taxonomic identification (Culley 2013). As the example in Suppl. material 1 shows, community ability to effectively link digital representations of voucher specimens with other data types, such as literature, people, genetic sequence information, traits, or even to assert and sustain semantic links between vouchers continues to be seriously hindered by the lack of PIDs and related services.

Persistent identifiers for specimens on the Internet can be compared in importance with digital object identifiers (DOI) for journal articles and datasets, where an article and its associated supplementary data is referenced via a unique DOI^®*5. Within scholarly publishing DOIs under the governance of the (International) DOI Foundation are having a transformational effect. Academic journal articles are more accessible and third-party services (such as those of Crossref) based on the growing PID graph are becoming more widely available and used (Cousijn et al. 2021). It is, for example, increasingly possible to link research outputs to the project grants and funding initiatives under which those were produced, as well as to work out the relations and collaboration among authors. In the future it will become possible to associate downstream impacts with the research outputs that created them, thus enabling measurements of value for taxpayers’ money. Similar effects are being seen in other industry sectors, such as film/tv and construction where DOIs are being adopted.

In collections-based science, an interconnected specimen graph based on PIDs can, in the future will reveal connections between specimens, and between specimens and the data derived from them or associated with them. An innovative range of exploratory, analytical and mining services is likely to emerge to exploit the graph, leading increasingly to industrialisation of research with machine-actionable metadata allowing for automated actions on data and flexible cross-domain data discovery and recombination.

The Handle System (Sun et al. 2003) is a PID scheme that is used widely, in many guises. As the basis for a persistent identifier mechanism for DiSSCo it is a consequence of the choice of the Digital Object Architecture as the architectural basis of DiSSCo data infrastructure (Hardisty et al. 2020). Digital objects of all kinds, including Digital Specimens need to be identified and they need a global resolution mechanism. The Handle System provides both of these things, whilst also being capable of both exhibiting and exploiting loose coupling with Web technologies.

The Handle System has several characteristics that are very desirable from the DiSSCo perspective:

• The namestrings used for the identifiers are location independent, making them immune to changes in the location of the identified object (provided the binding between the name and the location of the object is properly maintained);
• As pointers to objects, Handles are easily repairable when they break;
• There is transparency for identifying objects of any kind - both physical and digital;
• Identifiers are impervious to changes in underlying implementation technology over the curation timescales typical of the natural science collections sector (many decades, with a target of 100+ years); and,
• The Global Handle Registry infrastructure has reliability, robustness, and resilience for continuous distributed Handle resolution.

Thus, the Handle System and its multiple implementation options is the focus of the present appraisal.

The choice is further justified because FAIR Digital Objects (of which Digital Specimens are a kind) and their Handle-based PIDs are core elements of the European Open Science Cloud. Achieving compatibility with identifiers within the European Open Science Cloud (EOSC) for interdisciplinary interoperability and reusability is a further important benefit of investing in support for a DiSSCo PID mechanism.

Requirement: Scale for specimens, scale for machines, scale for performance, scale for global use.

Estimates suggest there are approximately 3 billion specimens in natural science collections around the world (Duckworth et al. 1993, Wheeler et al. 2012). Half of those are in Europe, housed in 120+ collection-holding institutions across more than 21 countries. These are used daily by more than 5,000 full-time scientists. Estimates suggest approximately 300 million specimens have been digitized to some degree to date in Europe, although not all that data is yet shared and publicly available.

When digitized, each resulting ‘Digital Specimen’ must be persistently and unambiguously identified. Subsequent events or transactions associated with the Digital Specimen, such as annotation and/or modification by a scientist must be recorded, stored and also unambiguously identified.

Suppl. material 2 provides an estimate of the number of PIDs likely to be needed throughout the DiSSCo lifetime and beyond. The numbers of identifiers that will be needed will rapidly surpass (by two orders of magnitude) the hundreds of millions of persistent identifiers already in use today. Thus, the scalability of a PID scheme to potentially handle several hundred billion identifiers is a very important requirement. At such levels, and with such large numbers of identifiers, machine-assisted services for working with PIDs will be essential.

A further important consideration is the way control can be exerted over the administration and use of the parts of the namespace hierarchy allocated for DiSSCo. Control and delegation of control over a large range of second or third-level name segments is likely to be needed. By subdividing control for administration and resolution purposes into smaller name segment blocks, better (i.e., faster, more balanced) performance can be achieved. High throughput workflows with multiple Digital Specimens as inputs and actionable steps that involve machine learning algorithms can rapidly generate tens of thousands of artifacts, each of which might need identifying; not forgetting that such workflows also require rapid resolution responses when manipulating intermediate artifacts.

Requirement: User confidence in the PID scheme, seeing it as appropriate to their needs and trustworthy.

Building trust doesn’t happen overnight or by accident. Trust is a result of specific actions that address consistency, quality and excellence. Being consistent with a PID scheme is about reliably and continuously sustaining services for creating, assigning and resolving PIDs over the long-term. A quality PID scheme (in the sense of a scheme that conforms to all the requirements stakeholders place on it) unambiguously delivers accurate and authentic data, traceable back to its points of origin and described by metadata adjusted to the specific community needs. Excellence in a PID scheme comes from being able to operate the scheme responsively, efficiently and cost effectively in a way that encourages convergence towards common digital scientific curation, publishing and access/use policies and practices.

Consistent, high-quality and excellent delivery of PID services built behind a branding that creates instant recognition for PIDs of the chosen scheme as the unambiguous pointers to specific accurate and authentic digital data about a specimen, including an unbreakable link to the corresponding physical specimen acts to confer authority. Trustworthiness should follow. Information quality is the strongest factor to influence organizational benefits through perceived usefulness and user satisfaction (Park et al. 2011).

Requirement: Many decades, more than 100 years.

An appropriate PID scheme must address the very long-term needs for durable identification in natural science collections, which can extend over many decades to 100 years or more. Specifically, the resolution component must be capable of accurately resolving PIDs and redirecting users and machines to the location of the data, even after new PIDs cease to be created and assigned. A maintenance responsibility and function for that must continue until such time as PIDs are no longer used for referencing digital specimen data. It’s clear that persistence goes hand in glove with governance.

Governance: An internationally acceptable governance mechanism by stakeholders themselves.

Responsibility for persistence lies ultimately with the community requiring it and, although there are several variations in the practical possibilities for maintaining guarantees of persistence, having a governing stake over the long-term for a chosen scheme is essential

DiSSCo is an international undertaking spanning multiple institutions across many countries, each with their national interest. A governance mechanism must be sensitive to local/national issues as well as fitting with the nature of DiSSCo as a European research infrastructure. Considering also the desirability and potential for global adoption of a single scheme (below) flexibility for extension to worldwide governance is an important dimension.

The surest way to deliver trustworthiness and persistence, and commitment to those aims is through governance by stakeholders themselves with accountability to the wider community. This is in the interests of both the stakeholders and the wider community. Despite that delivery of scalability might be best achieved through outsourcing and subcontracting, responsibility for that and its governance still rests with the stakeholders.

Requirement: PIDs appropriate to the digital object type being persistently identified.

The digital object to be identified and the circumstance of the object’s use dictate that PIDs should be appropriately chosen to reduce duplication, proliferation, and confusion of PID scheme types.

With differing circumstances of use and type specific metadata needs, DiSSCo has identified several categories of digital object, in addition to Digital Specimens that need to be identified (Table 1), with suggestions for possible PID schemes that could be adopted for each.

Requirement: Extensible towards a single PID scheme that could be adopted globally.

A single worldwide PID scheme for identifying digital specimen data is highly desirable because collections-based science is global, even if collections are both managed and used locally. It will become increasingly difficult for scientists when different PID schemes are used in different countries or regions.

A Digital Specimen PID scheme that fits with current regional and institutional practices for physical specimens and local specimen records is desirable and possible. The success and acceptance of a PID scheme for Digital Specimens will depend not only on the persistence of the PID layer but (as with DOIs) in a large degree also on connection to stable local resources and landing pages, such as provided by CETAF Stable Identifiers and International GeoSample Numbers (IGS). Neither CETAF Stable Identifiers nor IGSNs represent Digital Specimens. They are conceived as identifiers for physical objects in institutions (Güntsch et al. 2017, Lehnert et al. 2019). This role will continue to be important after a PID scheme for Digital Specimen data is established. The scheme should be usable to make references to specimens, in literature or elsewhere that have not yet been digitized. This can be done by creating a PID for an 'empty' Digital Specimen object even before digitization has commenced.

Global extension of a chosen PID scheme does not alter the principal requirements around scalability, persistence and trust. However, consideration for how a regional (e.g., DiSSCo) level governance mechanism can interact with or evolve towards a worldwide governance model is an important factor to consider.

The Digital Object Identifiers (DOI) scheme is the well-known PID scheme widely used in academic publishing, research and other sectors. In publishing it has become instantly recognisable through its use of the ‘10dot’ prefix (as in this example: doi: 10.1000/182 or its fuller DOI URL form https://doi.org/10.1000/182), its regular occurrence in article header information and bibliographies, and through marketing of the DOI brand. More recently, DOI is increasingly well known as an identifier for published datasets. Around 240 million DOIs have been registered to date though ten Registration Agencies (RA) that include Crossref, DataCite, EIDR and the Publications Office of the European Union. These RAs provide DOI registration, resolution and other services to their respective communities under different business models. The DOI PID scheme is governed and managed on behalf of its RA members by the DOI Foundation (IDF). Collectively the IDF and its RA members assume the long-term responsibility to maintain and sustain the DOI PID scheme for everyone.

An International GeoSample Number (IGSN) is an alphanumeric code, obtained through one of the IGSN Allocating Agents, for uniquely identifying material samples from the natural environment together with their related sampling features. Like schemes such as the Deep Carbon Observatory Identifier (DCO-ID), it is a specific case of the Handle System five-digit prefix scheme (below) in which the alphanumeric code forms the suffix of a Handle with prefix 10273. With roots in the identification of geological samples, the IGSN system assigns identifiers mainly for non-biological samples such as rock, mineral and fossil specimens, dredges, cores, etc. Small numbers of biological and archaeological samples are known to have been registered as well.

At the time of writing (March 2021), the IGSN e.V. implementation organisation and its stakeholder community is concluding a strategic review project funded by the Sloan Foundation. A plan and roadmap - ‘IGSN 2040’ - will guide IGSN towards an mature future as the global PID scheme for material samples of all kinds, not only geo/earth samples. The work intends to re-design and mature the existing organization and technical architecture of the IGSN scheme towards a global technical and organizational infrastructure - the Internet of Samples (Davies et al. 2021).

The European PID Consortium (ePIC) provides PID services for the European and wider international research community under the top-level Handle prefix “21dot”. The expectation is that the ePIC consortium will provide PID services to the European Open Science Cloud (EOSC) as a highly reliable, persistent and high performance service. Governed by a Memorandum of Understanding that aims to provide long term reliability, ePIC is a Europe-wide consortium of European information technology service centers for science. GWDG, one of the nine Handle system Multi-Primary Administrators (MPA) responsible for sustaining the Global Handle Registry Services is among the members and takes a leading role. The ePIC Consortium, especially by leadership of GWDG has been instrumental in devloping the persistent identifier policy and architecture for the EOSC (European Commission 2020, European Commission 2021a).

CNRI Inc. as the founder of the Handle System and one of its Multi-Primary Administrators (MPA) with shared responsibility for the Global Handle Registry offers Handle prefixes and registration services for a small annual fee. An organisation requiring PID services enters into a Registry Service Agreement to act as a Local Handle Service Provider (LHSP) operating and maintaining software systems providing local handle services (LHS) for minting, registering and resolving PIDs under their registered prefix(es).

Historically, five-digit prefixes of the form “1nnnn” have been registered. Already mentioned examples including prefixes allocated for IGSN (10273) and the Deep Carbon Observatory (11121) but there are many others. Whilst these legacy prefixes can remain in use in the future, new registrations of five-digit prefixes are deprecated in favour of the second-level and three-segment prefix schemes.

The second-level (or two-segment) prefix scheme is the general version of the more specific DOI and ePIC administration schemes described above. In this scheme, authorised Registration Agencies act for one of the nine credentialled MPAs to administer a portion of the namespace for specific PID user communities under the relevant top-level prefix.

Of the MPAs currently authorised, three have global scope and relevance for DiSSCo, namely: The DOI Foundation, GWDG and CNRI. Respectively, these MPAs are presently responsible for the 10dot, 21dot and 20dot top-level prefixes. The remaining six MPAs all operate with territorial scope in countries outside Europe and are not considered further.

Three-segment prefixes have replaced the deprecated five-digit prefix scheme (above). Three-segment prefixes allow a more granular level of prefix sub-division beyond second-level prefixes for administrative convenience to reflect organisational divisions and responsibilities. The current prefix allocation scheme operated by CNRI Inc. through their Handle.Net^® Registry is a two-segment/three-segment scheme i.e., both are possible under a top-level '20dot' prefix.

The three-segment prefix 20.5000.1025 is presently assigned to DiSSCo for experimental purposes. Since 2019, DiSSCo has been using this for evaluation and testing of the Digital Specimen Architecture approach and its technology components through the DiSSCo Natural Science Identifier Registry (NSIDR) demonstrator. Such prefixes are an attractive option as they offer a low-cost path of least resistance to piloting DiSSCo operational services in the short term.

At the highest level of namespace administration, DiSSCo could request a new distinctive two-digit top-level prefix specific for identifying Digital Specimen objects. This could sit alongside existing top-level prefixes such as those for DOI (10dot), general Handles (20dot) and European research (21dot) – to give just three examples.

Choosing this variant would require DiSSCo to adopt an operational mode of either allying with an existing MPA to obtain a delegated top-level prefix or to act as an MPA in DiSSCo's own right and be assigned a top-level prefix by the DONA Foundation, the non-profit body responsible for overseeing Handle System administration. The issues associated with adopting a new two-digit top-level prefix are complex – organisationally, financially and politically.

There is increasing evidence of emerging national-level support and services in several countries for persistent identification in the higher education and research sector. In part, this is due to the 'allocating agent' member model applied by RAs such as Crossref and DataCite. Much more so, it is due to recognition at the national level of the importance that PIDs play in connecting up the different parts of the research (information) ecosystem and as part of countries’ commitments to making publicly funded science openly accessible.

France, for example, in its National Plan for Open Science (MESRI. 2018) has declared national commitment to making open science a normal part of everyday practice for researchers; and to helping to define and regulate the building blocks of the open science ecosystem, such as Crossref and DataCite for DOIs and ORCID for researcher identifiers. This kind of declaration is typical in many countries.

In Germany, the ZBMED Information Centre for Life Sciences is a member of the DataCite consortium, taking the role of a DOI broker or allocating agent for German research institutions in life and medical sciences. The German National Library of Science and Technology (TIB) provides the underlying technical infrastructure and offers a similar service with focus areas in engineering and technology.

In the Netherlands, the cultural heritage and the library communities are active adopters and users of different types of persistent identifiers. The most common PID types in the Netherlands are DataCite DOI (operated through Delft University of Technology), the ePIC Handle system (through SURFsara) and URN:NBN (though the National Library and DANS). The National Coordination Point for Research Data Management organizes various workshops and training focusing on different aspects of research data management.

In the UK, Jisc is working to select and promote a range of unique identifiers in collaboration with relevant partner organisations and funders of research, who may (for example) consider mandating the use of such PIDs as a condition of research grants.

In many of these cases, the effort is mainly directed towards identifying datasets, people, organisations, research grants and research outputs. Only the UK's HeritagePIDs project, as far as we are aware is investigating approaches to PIDs for institutions across the UK heritage sector as the sector considers how to work towards an open national collection.

As a Europe-wide research infrastructure with a new requirement to persistently identify Digital Specimens, DiSSCo will be best positioned by selecting and promoting identifiers with relevant partner organisations on the European rather than national level. For other kinds of identifiers (DOI, ORCID Id, GRID/ROR) it makes little difference whether these are provided/used on European or national level. This is the matter of operationalizing a PID scheme, explored further in the following section.

Within the overall category of choosing to administer a root or top-level name segment there are two options: to ally with an existing MPA or to become and act as an MPA.

There are presently nine MPAs authorised by the DONA Foundation, collectively sharing the responsibility to operate the top-level Global Handle Registry (GHR). Six of these are territory oriented and can be discounted for alliance purposes whilst the remaining three (IDF, GWDG, CNRI - see Table 2) have non-territorial global scope and are worth considering.

Within the overall category of choosing to administer a lower-level name segment under a top-level name segment there are three options: to use the services of an existing RA, to ally with an existing RA or to become/establish a new RA.

A Registration Agency provides services to those wanting to register PIDs. These include allocation of prefixes within the range administered by the RA, registration of PIDs and capture/maintenance of associated metadata.

The RA scenarios (A5, C5, E5, F5) are attractive on several levels. They offer the promise of being able to develop and acquire the trust of the target community through recognition and confidence in a relevant RA market brand. This can be branding created by DiSSCo itself along the lines that RAs like EIDR and BSi.Identify have achieved or branding in combination with the brand reputation of the top-level prefix administrator (DOI Foundation, ePIC consortium, Handle.net). To what extent is it helpful to leverage an existing brand versus developing an own brand such as 'Natural Science Identifier (NSId), for example? DOI is a very strong trademarked brand. ‘Driven by DOI’ or ‘DOI powered’, for example can be an enticing strapline that conveys confidence that DiSSCo is not doing much different or out of step with other, well established uses of the DOI brand. The ePIC brand is likely to grow in the context of the EOSC. ePIC services are not only European but global in scope and eventually will be tailored specifically towards the research (data and infrastructures) communities. Handle.net^® is widely used but is better known in technical circles than non-technical ones. Brand fit has to be considered and there is no doubt there is a good fit with DOI even if DiSSCo takes its own branding route too. “Natural Science Identifiers, driven by DOI” has a reassuring confidence about it, suggesting that DiSSCo is buying into an established mechanism whilst offering the custom characteristics needed by the sector.*8 Operational compatibility with the journal and data publishing businesses, as well as with EOSC is assured; as are persistence and sustainability through the IDF with the support of its RA members, CNRI and the industry sectors that rely on DOIs now.

On the other hand, DiSSCo should remain aware of the increasing importance of DOIs in and to the commercial sector. There is a potential risk to IDF and its non-profit RA members of increasingly coming under commercial influence that can grow stronger in the future. That might be an argument for a different scenario to avoid that the publicly-funded heritage collections experience pressures that are disadvantageous. This could push DiSSCo more in the direction of ePIC or Handle.net if that is thought to be a problem. Or it could dictate that as well as acting as an RA, DiSSCo should aim to ensure a strong alliance and influence in IDF.

Handle.net is a weaker brand model, sustained by CNRI and its business. RAS (Reliability, Availability, Serviceability) policies and plans are unclear although CNRI is known to be working on this. CNRI Inc., is a privately held, USA-based not-for-profit organisation established to “undertake, foster, and promote research in the public interest.” As the founders of the Handle system and operator of both local Handle registry services (Handle.net®) and a significant part ofthe Global Handle Registry as well, CNRI has a large stake in play. CNRI also offers subcontract technical and operational services to some of the other players, including operating the DOI infrastructure on behalf of IDF. CNRI’s ‘HDL enabled’ logo and strapline are not as well known as the DOI brand but technical capability is strong and, subjectively, resilience seems reasonable. It is important to note that the DOI system uses Handle.net as a necessary but not the only component for the DOI system.*9 CNRI thus is sustained through the DOI ecosystem as well.

With its strapline of ‘persistent identifiers for eResearch’, the European PID Consortium (ePIC) is backed by some of the strongest research sector players in Europe, including CSC (Finland), KTH (Germany) and SURFSara (Netherlands), as well as the Swiss National Computer Centre, the German Climate Research Centre and the Greek Research and Technology Network. ePIC will probably gain more members and strength as EOSC becomes more established. Ultimately, it will become a reliable service provider with a specific research community focus.

Could DiSSCo as an entity live with working under an existing top-level prefix (10dot or 21dot)? The current consensus is that DiSSCo could do so, especially if DiSSCo were to establish its own distinct RA. That model has clearly been successful for Movielabs with EIDR, which is completely differentiated from anything DataCite and Crossref do. There’s no confusion because EIDR is the brand and the users don’t care that it’s 10dot or that 10dot is also in use in other sectors. NSId as a 10dot brand isn’t unappealing, perhaps with a distinctive second-level prefix such as '10.22/' or '10.30/' for example. This is a solution to take advantage of the DOI ecosystem and yet still allows DiSSCo to create its own unique service space. And that might be a deciding factor for the IDF in preference to GWDG/ePIC or Handle.net, whose top-level prefixes 21dot and 20dot and associated service spaces are not so well developed as yet.

An important concern for the present options appraisal is to understand the opportunities presented by naming and adopting a community specific identifier like 'Natural Science Identifier' (NSId), the value to assign to these and the implications arising in terms of commitments to ensuring that digital specimens on the Internet remain openly findable, accessible, interoperable and reusable over many decades (100 years’ time). How DiSSCo and its counterparts globally opts to play in the Handle System is ultimately a strategic decision about the value and branding of digital identifiers for digital specimens on the Internet. How does the European and worldwide natural science collections community want to be perceived in an increasingly digital future?

Community specific identifiers and digital identifier systems more generally do play an essential role by offering a focus on uniqueness and persistent identification of digital objects that leads to stable, authoritative packages of scientific information and trust. On the other hand they are the face of a resolution mechanism for instant and online access that makes (with software assistance) navigation and interactive use of assets easy. Assigning identifiers to digital artefacts and associating relevant metadata with such identifiers has been shown to substantially increase findability and use of artefacts (Khedmatgozar and Alipour-Hafezi 2017). This is in keeping with the tenets of the widely promoted and modern movement that aims to make open science a normal part of everyday practice for researchers and to make scientific research results open to all – researchers, companies, citizens.

Transversely, a strong role is placed on digital identifiers as enablers of machine-actionable support to humans in an age of Internet diffusion, workflow automation, data explosion and machine learning/AI. This has long been the case in software systems of all kinds using unique identifiers internally – much of the time hidden from direct view. This is a practice that dates to the 1980s in origin with the invention of graphical workstations. Only now do the descendants of such internal identifiers assume a much more prominent and public role thanks to the Internet, digital transformation and the need to manage – as opposed to just see and move – data on the Internet.

Digital identifiers are enablers for social and professional change in the way in which collection-holding institutions both manage and share their holdings and expedite the related data into the hands of users. Thus, while perhaps not being able to assign a quantitative value to identifiers, we can say qualitatively they already have a high value now. Coupled with transforming physical specimens’ data to digital specimens on the Internet, they have far more potential in coming years.

The question for DiSSCo and global counterparts is: How important is it to have a clear brand association between the objects identified and the identifiers themselves? What does that look like? What does it mean in terms of trust and persistence (and ultimately, governance) for a chosen PID scheme and its mode of operation?

Assigning value in terms of trust and persistence has a consequence both for the choice of PID scheme and for the choice of an operational mode within the selected PID scheme. The two choices are not completely independent of one another. How does such value need to translate to a clear brand association between the objects identified and the identifiers themselves? And, discussed above how important is brand value?

That must be thought about in the context of the future value chains founded on natural science collections, especially those in recently described notions of extended specimens (Webster 2017), next generation collections (Schindel and Cook 2018) and the role of collections in ensuring critical research and education in the current century (National Academies of Sciences, Engineering, and Medicine 2020).

‘Open Digital Extended Specimens on the Internet’ (openDS) enhance the value chains founded in natural science collections. These value chains extend from initial gathering and organisation of specimens, through conduct and commercialization of specific science based on specimens, to sharing ensuing economic and social benefits in a fair and equitable way. Products of digital value chains can provide the evidence for regulatory processes in health, food, security, sustainability and environmental change, and new educational uses. Future software applications can work with and on Digital Specimen objects to provide more sophisticated computer-assistance to both the present day known work tasks and to unimaginable future works of collection specialists, scientists and others working daily with specimens. Natural science collections and the science and other values that flow from them are part of the common good that must be shared with everyone. Open digital specimens on the Internet, persistently identified in a manner that makes them instantly recognisable to end-users – researchers, companies, citizens – are a specific category of raw materials for new knowledge generation (research) and acquisition (teaching and learning).

In consideration of the foregoing, the strongest option across the studied major dimensions of the available Handle System PID schemes and operational modes is for DiSSCo to use DOIs to identify Digital Specimens. The case for choosing DOI comes out slightly more strongly than choosing ePIC for reasons related to the substantial achievements, operational experience and reputation of DOI/ IDF to date. Operating under another Handle-system prefix than those used by IDF and ePIC is the substantially weakest option because of the difficulties associated with introducing an identifier that is not perceived to be a DOI. The term ‘DOI’ is trademarked by the IDF and thus not available for describing other identifiers.

The practical and sensible avenue to explore further are the options to establish and become an RA member of the DOI Foundation (option A5) and to enter a strategic alliance at the level of the DOI Foundation (option A1). These options are likely most effective when actioned in combination.

Although DiSSCo is a European endeavour for digital unification of European natural science assets, the present options appraisal should also consider relevant developments from elsewhere, especially outside of Europe noting again that one of the main requirements for DiSSCo's preferred PID scheme is potential for global adoption. Here follows a dissection of relevant facts.

GBIF has been using DOIs for identifying occurrence datasets and queries for many years. GBIF is investigating what is needed to persistently identify each individual occurrence record in a resolvable manner. The Darwin Core term ‘occurrenceID’ is already an identifier of each individual occurrence but in many cases this is not fully and directly resolvable back to the original data about that occurrence. Identifying an occurrence with a DOI or IGSN could solve that problem. GBIF has been investigating both schemes.

International Geo Sample Numbers (IGSN) have gained traction in the geoscience community. The iSamples Research Coordination Network (RCN), 2014 – 2019 was successful in laying the groundwork for the expansion of the IGSN scheme to other domains where material samples are of importance, such as natural and environmental sciences, material sciences, agriculture, physical anthropology, archaeology and biomedicine. In June 2020 a follow-on iSamples project received funding to build infrastructure to begin this expansion and engage new communities over the coming three years.

The Arctos collection management system (CMS), which curates and serves data from 180 separate collections can allocate DOIs to specimens recorded in that database.*10

The Specify consortium may decide to add functionality to the Specify collection management system (CMS) to begin adding DOI prefixes or IGSN prefixes in front the GUIDs that Specify CMS already generates for each specimen record. That would probably be easy for them directly (or with a collaborating third-party) to implement and deploy e.g., with a ‘PID registration plugin’ such as the IGSN compatible plug-in (“iSamples-in-a-box”) that will be developed in the previously mentioned iSamples project. Apart from installing the plugin, a collection manager would only need to sign up with a registration agency such as DataCite to begin registering DOIs.*11 The cost of that would be perhaps $5-10k per year to an institution. Specify is one of the most widely used CMS software solutions, especially in North America. Presently there are more than 275 installations of the software, managing 450 collections across 38 countries.

The CMS landscape is fragmented. In a survey by DiSSCo in 2017 (Casino et al. 2019) it was found that many institutions use more than one collection management solution. 117 systems were reported by 89 institutions. Often, in-house custom solutions (36%) or solutions based on MS Access, MS Excel, MS Word or Filemaker (21%) are used. The top 3 CMS found were Specify (7%), Adlib (3.5%) and JACQ (3.5%). In a recent survey (Spring 2020) of 200+ respondents presented at the 2020 virtual conference organised by the Society for the Preservation of Natural History Collections (SPNHC) (SPNHC & ICOM NATHIST 2020), the top three CMS in use were found to be: Specify (23%), Axiell EMu (13%) and Arctos (12%), followed by ‘Microsoft Access/Excel & Filemaker’ (25%). Custom and minority solutions account for the remainder. That survey was mostly an anonymous straw poll so the responses were technically international; but for the fifty or so people contacted after completing the survey, most seemed to be from North American institutions. Thus, it's likely there's North American bias in the result. Nevertheless, taken all together the results illustration that multiple different CMS softwares are in use and the domain is ripe for churn for new capabilities.

Despite that only one DiSSCo member institution makes use of Axiell EMu and that Arctos is not used by any institution in DiSSCo, what Specify, Arctos and EMu decide to implement in terms of PIDs for specimens could have an important bearing on directions in general and thus for DiSSCo and global counterparts. If DOI and/or IGSN enhancement in CMSs were to go ahead, that would be quite impactful. The main iDigBio players use mainly Specify so this would quite suit them and probably VertNET as well.

Lastly, it is important to recall in this context the similarities and differences of DiSSCo’s approach for the digital representation of physical specimens (Hardisty 2019, Hardisty et al. 2020) with the Extended Specimen Network (ESN) concept (BCoN 2019, Lendemer et al. 2019). In the ESN concept presently, digital representations originate and root in the CMS record itself, adding external resources to that. Both Arctos and Specify CMS are at least partially capable of making those linkages to external sources and that capability can grow. But this leads to more complexity within and across such CMSs. Extensions and harmonisations are in the hands of the different CMS vendors and development consortia. Enhancing Specify records with a Handle prefix (as already done by Arctos) is already then persistently and uniquely identifying extended specimens. A PID per specimen record could be registered, made up of a prefix and the GUID that Specify already automatically generates for each record. DiSSCo’s Digital Specimen concept is different here, with each DS as a digital entity distinct and separate from (external to) to a specific CMS record. This involves a separate/new PID to identify the DS alongside, for example the CETAF Stable Identifier identifying the physical specimen and its corresponding collection management database record. The DS and the physical specimen remain coupled through the information associated with the PID of the DS. CMSs like Specify and Arctos can easily be adapted to point in the other direction to corresponding DSs as well. Discussions to achieve technical convergence, as we explained earlier are in progress*1 with the expectation that this will be achieved. 'Digital Extended Specimens' will be distinct, identified digital entities that represent specimens on the Internet, extending the information about them normally held in institutional collection managment systems. They will be separately processable.

In the simplest sense, offering a PID service means providing, on the one hand a resolver that provides a table lookup with redirection to the asset of interest, and on the other hand a mechanism for capturing information into that table to make the resolution possible. However, over the years, the global scientific community, data and service providers realised providing such service is not such a simple endeavor. Well-founded PID initiatives have focused on building robust, trustworthy, reliable and sustainable service that can cater to the global research community.

DiSSCo needs to warm up its engines to begin delivering pilot and pre-production level services whilst at the same time consulting with global counterparts to develop the necessary governance, finance, operations and architecture that can ultimately lead to a worldwide PID scheme for Digital Extended Specimens.

In the time prior to a DiSSCo RA becoming operational (c. 18 months from decision), DiSSCo will join IDF as a member and work to pilot the simplest PID service whilst developing the four GOFA areas (governance, operations, finance, architecture) and processes as prelude to an RA member application procedure to the IDF. This involves developing a service model that fits the DiSSCo service portfolio guidelines with key performance indicators to measure success. Preliminary elements of some of this necessary work are mentioned below.