The PID Guide is meant to help you learn and think about important PID topics, and guides your first steps towards selecting a PID system.
The Guide focuses on the Dutch Digital Heritage sector, with libraries, archives and museums, but also research organisations in mind.
The PID systems included in the Guide are available for and used in this sector, and meant to identify cultural heritage objects. Not people, grants, organisations et cetera.
By checking a position's "Extra importance?" checkbox, you indicate that that position's topic is extra important for your situation and decision-making. That position's score will weigh double. You can check a maximum of five boxes.
PID's are a managed resource. They will not magically keep referring to your objects if you change the objects' locations (URLs). You will have to keep your PIDs up to date.
Glossary:
Landing page: a webpage that shows the current location of the object and provides metadata like object details and information on any access restrictions.
Metadata: data about data, answering questions as 'who', 'what', 'where', 'when' and 'why'.
Persistent Identifier (PID): a unique identification code attached to a digital object and registered at an agreed location.
PID (service) provider: an organisation providing services related to a PID system, such as the ARK Alliance for ARK, TU Delft Library (DataCite Netherlands) for DataCite DOI, SURF for the EPIC Handle System and the KB, National Library for URN:NBN.
PID system: the PID systems included in this Guide are ARK, DataCite DOI, Handle System and URN:NBN. For a longer list, see Overview of PID systems elsewhere on this website.
Prefix and suffix: parts of a PID, where the former usually identifies the owner of the PID and the latter further identifies the object. In https://hdl.handle.net/10648/6175fdf5-dc90-4b71-9922-26223b00c432, 10648 identifies the National Archives of the Netherlands and 6175fdf5-dc90-4b71-9922-26223b00c432 identifies the National Archives of the Netherlands' inventory of the archives of the Van Oldenbarnevelt family. The (suffix) strings that form identifiers, regardless of type (ARK, DataCite DOI, Handle or URN:NBN), can be generated by a variety of means, from simple counters, to human-invented semantically-laden names, to 37-character opaque UUIDs, and more.
Resolver: an application that allows the user to determine the location of an object, based on a PID. In the National Archives of the Netherlands PID example, https://hdl.handle.net is the resolver. The combination of resolver and PID makes a PID actionable.
Tombstone: a landing page with information about a deleted object, e.g. explaining when and why it had to be deleted.
THEME 1 Aims: What do you want to achieve by implementing Persistent Identifiers?
Position 1
I want a PID system that comes with its own metadata policies and requirements.
Some PID systems come with metadata policies and requirements, e.g. regarding required metadata fields.
Answer
The more freedom you require in terms of policy creation, the more internal regulation you will have to govern and maintain. URN:NBNs and DataCite DOIs have clear (and strict) policies, while choosing ARKs and Handles means you are quite free to create your own policies and apply them as strictly as you want.
Position 2
I want a PID system that comes with its own persistence commitment policies.
The more freedom you require in terms of policy creation, the more internal regulation you will have to govern and maintain. URN:NBNs and DataCite DOIs have clear (and strict) policies, while choosing Handles means you are quite free to create your own policies and apply them as strictly as you want. The ARK philosophy that persistence is not binary allows providers to return a statement describing their persistence commitment along with object metadata in order to set expectations about what may change over time.
Position 3
My Persistent Identifiers are in place so that humans can cite and reference objects.
Who is the audience for your PIDs: human or machine? Objects can be materials, works, databases, table cells, images, image tiles, data sets, linked data concepts, etc.
Answer
DataCite DOIs are designed for reference and citation of research data and research outputs, and are known and accepted as good practice in research communities. URN:NBNs are primarily designed for identification, and less so for citation purposes. Handles are an all purpose persistent identification system and generally useful assigning identifiers to a large number of objects. ARKs are also designed for the reference and citation of any object type, whether digital, physical, or conceptual.
Position 4
My Persistent Identifiers will be used and transcribed (copied) more by machines than by humans.
For example with aggregation services such as Europeana or in APIs.
Answer
If the main goal is that aggregators such as Europeana and Digitale Collectie Nederland can always track and find your objects, then Handle is probably the best choice for you. You can also use DataCite DOIs and URN:NBNs, however these focus more on citation than the machine readable aspects. ARKs can point either to a landing page or directly to machine-processable data, such as a spreadsheet or an inline image.
Position 5
I want to use my Persistent Identifiers in all stages of research projects and processes.
Including intermediate and/or temporary results, not just for publishing final results.
Answer
Whenever you aim to link Persistent Identifiers to stages in the input or output, or to various steps in your data workflows, then you aim to make your research as transparent and reproducible as possible. Handles lend themselves to this type of process. DOIs and URN:NBNs are better linked to finished, published datasets. ARKs are meant to reference objects at any point in their lifecycle, from an object that is planned but doesn't yet exist, to an embryonic object that has no metadata yet, to an object under active development and review, through to formal release, to revision, to deprecation, to tombstone.
Position 6
My Persistent Identifiers will point to objects in a Long Term Digital Preservation Depot.
The depot must aim to provide long term sustainable access to digital objects.
THEME 2 Context: What type of objects are you creating Persistent Identifiers for?
Please note: position 7-10 are about objects, position 11-14 about sets, collections and other aggregations.
Position 7
I will predominantly use my PIDs for objects that are books, journals, newspapers and magazines.
Some PID systems lend themselves better for certain types of objects, or are more used in communities linked to those types of objects.
Answer
With URN:NBNs the emphasis is on the sustainability of publications, data(sets) and accompanying metadata. A URN:NBN is therefore ideal to link to these types of publications. DataCite DOIs are, as the name suggests, generally suited to pointing to (scientific) data and datasets. Handles can be applied to individual objects (e.g. files or scans). ARKs are used as a lower-cost option for small publishers (e.g. OJS-based journals) as well as for scans of historic newspapers.
Position 8
I will predominantly use my PIDs for physical objects, such as museum artefacts, photos or books.
Some PID systems lend themselves better for referencing physical objects.
Answer
DataCite DOIs support the identification of physical objects. Handles are used by many cultural heritage orgnisations in the Netherlands in conjunction with their collection management system to track their collection items. DataCite DOIS and URN:NBNs require a landing page with accompanying metadata. Internationally, millions of ARKs are assigned to physical objects. ARKs are also gaining track in the Netherlands, see e.g. the list of Dutch ARK users elsewhere on this website.
Position 9
I want to use my PIDs to point directly to objects, not metadata or landing pages.
Some PID systems require that PIDs point to metadata or landing pages.
Answer
DataCite DOIs require mandatory metadata elements. URN:NBNs point to a landing page where you can publish or edit the metadata for an object. Handles do not require specific metadata, though mandatory and optional fields can be added. Handle PID provisioning services such as EPIC at SURF may however limit the number of metadata fields as they are not meant to act as a metadata server. With all these PID systems, the Persistent Identifier can always direct to the metadata record of an object if required. While ARKs support objects under development, from no metadata to rich, unconstrained metadata, it is encouraged that as early as feasible, object metadata be able to supply the three Dublin Kernel elements (who, what, when).
Position 10
My Persistent Identifiers point predominately to individual, digitised objects such as letters, maps or charters.
Do you want to point to individual and mainly scanned objects, or to data(sets) or metadata?
Answer
With URN:NBNs the emphasis is on the sustainability of publication, data(sets) and accompanying metadata. A URN:NBN is therefore ideal to link to publications. DataCite DOIs, as the name suggests, generally point to data and datasets. Handles and ARKs can be applied to individual objects (e.g. files or scans) without stopping at a landing page.
Please note: position 7-10 are about objects, position 11-14 about sets, collections and other aggregations.
Position 11
My Persistent Identifiers point predominantly to collections of objects, rather than to individual objects.
E.g. collections of books, artefacts or datasets.
Answer
With URN:NBNs the emphasis is on the sustainability of the publication, data(sets) and accompanying metadata, but this makes URN:NBNs less suitable as a method for the persistent identification of collections. DataCite DOIs, Handles and ARKs are more suited for this purpose.
Position 12
I want to be able to change (the content of) my objects, which have been assigned a Persistent Identifier, without changing the Persistent Identifier itself.
What if you want to edit or change the object without creating a completely new version and Persistent Identifier?
Answer
DataCite DOIs and URN:NBNs in principle require objects that stay the same. Both DataCite DOIs and URN:NBNs allow the object to be modified, but then a new (version) identifier must be created for the modified object. Handles and ARKs have no such requirements. So if you want to create different versions of objects while using the same Persistent Identifiers, then Handles and ARKs offer you this possibility, DataCite DOIs only partially and URN:NBNs do not offer this option at all. ARKs recommend that providers return persistence commitment statements that characterise the kinds of change to be expected before an object must receive a new identifier, e.g., corrections only, security patches only, appended content only. Please note that how to deal with versions of (meta)data is for a large part a decision you will have to make and communicate yourself.
Position 13
I want to be able to change (the content of) my metadata or landing page, which has been assigned a Persistent Identifier, without changing the Persistent Identifier itself.
What if you want to edit or improve the metadata or landing page, without creating a completely new version and Persistent Identifier?
Answer
URN:NBN and DataCite DOI require that what they refer to remains the same, see also position 12. With DataCite DOI, changing the metadata or landing page is possible. With Handles and ARKs you are free to make your own policy. Note that how to deal with versions of (meta)data is for a large part a decision you will have to make and communicate yourself.
Position 14
I find it important that the Persistent Identifier system requires specific metadata, so that the objects are findable in a standardised way.
Some Persistent Identifier systems require specific metadata.
Answer
DataCite DOIs require specific metadata to make the object findable. Connections to other related research outputs, people and organisations are made based on relationships and PIDs in the metadata. URN:NBNs point to a landing page where you can publish your metadata, and the KB, National Library expects you to provide this metadata. Handles do not require specific metadata (but you can mandate any type of metadata that you would like to). ARK providers are encouraged to supply at least the three Dublin Core Kernel elements who, what and when in order to support broad interoperation across object types and domains.
THEME 3 Use: How will your organisation make use of Persistent Identifiers?
Position 15
I want to be able to delete my Persistent Identifiers, when I delete what they point to.
Some Persistent Identifier systems allow the deletion of Persistent Identifiers.
Answer
If you want to be able to delete your Persistent Identifiers, you should be able to justify this decision. If you then still want to be able to delete your Persistent Identifiers yourself, then please consider if this is really what you want from this service. URN:NBNs do not offer an easy option for deleting Persistent Identifiers. The service provider must do this for you. With Handles, an administrator can delete your Persistent Identifiers. DataCite DOIs cannot be deleted at all, and the DOI must resolve to a landing page explaining why the object was deleted. These pages are usually referred to as tombstones. ARKs that have been publicised or that have unintentionally “escaped into the wild” are strongly encouraged to be deleted only if they continue to resolve to a tombstone record explaining what the original binding was and the reason for deletion. All other ARKs are considered private and experimental and can be deleted without ceremony.
Position 16
My organisation doet not have the IT capacity to implement Persistent Identifiers in our systems (e.g. collection registration system, website or repository system).
How much IT capacity (knowledge and time) does your organisation have at its disposal to implement Persistent Identifiers in your systems?
Answer
URN:NBN is a free service, but requires that you have a PID generator and deliver your PIDs to the resolver. For DataCite DOIs (technical) support and guidance is provided, but organisations must implement the DOI registration themselves. Handles have similar organisations providing services, e.g. EPIC at SURF. Various vendors of collection registration systems have implemented Persistent Identifiers in their systems (for e.g. Handle or ARK). You can also choose to implement and manage your own Handle server, but that requires more IT capacity. Apart from a few services that provide ARK support in various domains, some technical capability (e.g., installing an OJS or Omeka plugin) is required to implement ARKs at your organisation’s website. Thus, to implement Handles or ARKs, your organisation will need to have the appropriate technical capabilities at its disposal (for example, to run a Handle server or to install the open source EZID software).
Position 17
I am not prepared to pay a fee for membership or subscription to and use of a Persistent Identifier system (excluding implementation and additional service costs).
Some Persistent Identifier systems are free to use, others come with a fee.
Answer
For DataCite DOIs, a fixed organisation fee of ~500 euros per year is charged (for which multiple prefixes may be used), plus a DOI service fee for the number of registered DOIs. See the DataCite DOI fee model. EPIC Handles are ~50 dollars for registration (once) and then a yearly contribution of ~450 euros for the PID hosting service and ~50 euros for Handle prefix preservation. See the SURF Services and Rates brochure. If you were to host your own Handle server, you would have to compare the PID hosting service fee to your organisation’s hosting costs. The KB, National Library provides URN:NBNs as a free service. There are no fees to use ARKs, either per-ARK or to register as an ARK-assigning organisation.
THEME 4 Support: What do you expect from a Persistent Identifier Provider and their organisation?
Position 18
I find it important to be able to choose between different Persistent Identifier providers with my chosen type of Persistent Identifier system.
How much choice in providers do you require?
Answer
DataCite DOIs and URN:NBNs have national registration agencies. In the Netherlands this is for DataCiteDOIs TU Delft Library (DataCite Netherlands) and for URN:NBN the KB, National Library respectively. Handles and ARKs give you more freedom. If you are looking for a Handle provider you can choose from SURF’s EPIC service (Dutch based), another European EPIC provider (e.g. CSC, grnet, SND,etc.) or implement your own server. Currently most ARK providers modify their web servers to add ARK support.
Position 19
I find it important to be able to move to another Persistent Identifier provider without changing my chosen type of Persistent Identifier system.
How easy should it be to move to another PID provider?
Answer
Dutch organisations can obtain DataCite DOIs through the TU Delft Library (DataCite Netherlands), or become a direct member of DataCite themselves. Dutch URN:NBNs are taken care of through the KB, National Library only. Handles give you more options, as you can connect with (inter)national providers or implement and manage your own Handle server. ARKs are not recorded centrally, so changing to a new provider requires a one-line update to the ARK organisational registry. Currently (2023) there is not really a fully developed solution for moving your Persistent Identifiers from one Persistent Identifier system (e.g. Handle) to another (e.g. URN:NBN).
Position 20
I find it important that I can get support from a Persistent Identifier provider when my organisation is implementing Persistent Identifiers.
Do you want to be able to call a PID help desk and get technical support?
Answer
As the registration agency for DataCite DOIs, TU Delft Library (DataCite Netherlands) offers extensive support. The KB, National Library offers limited support in their capacity as registration agency for URN:NBNs. In the case of Handles, you need to do most of the work yourself, however SURF and EPIC can offer some level of support. Organisations that offer ARKs do not rely on a centralised external server (e.g., as DataCite does), so support is defined entirely by your service level agreement with your provider.
Position 21
I find it important that I can get support from a Persistent Identifier provider in the daily use and maintenance of my Persistent Identifiers, e.g. when I change the location of my objects and need to update my PIDs.
What level of support do you require once PIDs have been implemented in your systems?
Answer
Support for Handles is available through PID-providers such as EPIC at SURF. Or you can organize the use and maintenance yourself. With URN:NBNs you submit your changes to the metadata via your Registration Agency at the KB, National Library). For DataCite DOIs, organisations must themselves change the metadata or the location of their objects in the DOI registration system (via the user interface or API). ARKs are not recorded centrally, so implementing most changes requires no coordination. (All identifier types require centralised changes for rare events, such as when one organisation splits or merges with another.)
Position 22
I find it important that there are relevant statistics about the use of my Persistent Identifiers available, e.g. how often they are resolved.
Are (usage) statistics a feature of the PID service you are looking for?
Answer
DataCite DOIs offer the most possibilities and works on providing citation statistics. ARKs, Handles and URN:NBNs do not provide this as a standard service.
THEME 5 Technical: What do you expect from the Persistent Identifier infrastructure (in terms of technical quality) and the technical basis of the resolution of identifiers?
Position 23
My chosen Persistent Identifier system needs to be scalable, so that my PIDs are discoverable in a global context.
Not all Persistent Identifier systems provide a global resolver. (keyword: resolution)
Answer
URN:NBNs have servers that resolve locally and nationally. DataCite DOIs and Handles use centralised global resolution. ARKs can be resolved both locally and through the global ARK resolver at n2t.net.
Position 24
My Persistent Identifiers must point to an online address, or have a web address even if the object consists only of metadata (e.g. your Persistent Identifiers will never directly point to a physical object).
Do you want all your PIDs to point to something that is available online? (keyword: actionable)
Answer
DataCite DOIs are, as the name suggests, ideal for directing to datasets. DOIs should resolve to a landing page with information (metadata) about the object and how to access the object. URN:NBNs can point to physical objects such as books, but it is expected that there will always be a landing page with information about said object available online. Handles can be actionable Persistent Identifiers, but they can also point to physical objects without a web address. It is strongly recommended that, if mature enough to be publicised, ARKs resolve to digital objects or, for physical or conceptual objects, to landing pages with descriptions.
Position 25
If I have assigned a Persistent Identifier to an object, I also want it to be possible to point to a part of an object, e.g chapters in a book, or to a distinct webpage on a given (archived) website, without creating any additional PIDs.
Some PID systems provide features to point directly to parts of the object a PID points to. (keyword: part identifiers)