Object Persistence: a Matter of Service

I recently encountered the Archival Resource Key (ARK) identifier scheme and read its latest draft specification.¹ Its thesis is that the persistence – i.e., long-term identification and access – of any information resource – i.e., an “object” – is a matter of service.

A stewardship promise must come from a current provider, which may not be the original assigner, because an assigner simply cannot ensure persistence – either via object structure or via a naming scheme. The ARK spec makes a distinction between the so-called “memory institution” that acts as a name assigning authority (NAA) and thus associates an identifier with an object, and a service provider – a name mapping authority (NMA) – that actually stewards that information.

Without a description – “a record bearing witness to the identifier’s association with the object, as recorded by a set of object characteristics”¹ – real identification is incomplete. A naked identifier isn’t enough – you need the record, the “receipt.” The name mapping authority – which may be the same as the name assigning authority for as long as funding/politics allows – acts as host, as steward.

Persistent identification (naked identifier plus description record) is separable from persistent access. Description-only NMAs can redirect object-access requests to other NMAs, e.g. publishers that restrict access to object content.

A key function of ARK’s abstract model is for a service to deliver not only a descriptive metadata record, but to deliver also a policy declaration: a declaration of permanence for each serviced object – its availability, identifier validity, content invariance, and change history. NLM “permanence levels” are an example of such declarations.

On a practical level, ARK facilitates the distinction between name mapping and name assigning via the anatomy of a typical ARK:

https://<nma>/ark:<naan>/<assigned_name>

The above is highly simplified. <naan> is the name assigning authority (NAA) number, i.e. the the opaque number assigned to the NAA. <nma> is the name mapping authority – n2t.net is commonly used, just like doi.org is commonly used for resolving DOIs. https:// is common, though ARK is an abstract model with no concretions around use of e.g. JSON or even HTTP.² Finally, <assigned_name> can reflect object hierarchies and qualifiers for variants (versions, formats, etc.), all in ways formally given in the ARK spec.

A memory institution names objects with identifiers like ark:12345/x2fr5, and any number of NMAs may serve access to the named objects and/or to their descriptions. Just like your web browser discovers an IP address for a given domain name, your browser can discover one or more NMAs servicing a NAAN namespace and thus an ARK. This can work even if a preferred NMA is given but is not available, e.g. the ARK is presented as https://example.org/ark:12345/x2fr5.

ARKs are kind of like DOIs, but you can assign names freely under your assigned NAAN, you can migrate your object repository to any NMA at any time, and NMAs can be explicit to users about multi-faceted persistence policies on a per-object basis, all while guaranteeing permanence of the “naked” (string) identifier.