How to trust the persistence of GUIDs

After some rather quiet months, members of the Taxonomic Database Working Group (TDWG) have started to discuss the GUID (LSID) issue again.
Beneath the never ending discussion on which identifier system (DOI, LSID, PURL, handle etc.) is best, some other interesting issues have been raised: does a GUID system need central services, does it need a business model (money) etc.

Most interesting Roger Hyam has questioned the trustworthiness, thus persistence of GUIDs. He comments:

What I find most interesting is that the only people who promise persistence are people who are trying to persuade you to use their system. The basic message is “don’t trust them trust us”
....
When we say GUID persistence what we really mean is reliability of GUID resolution in the short term. This is something entirely different from persistence for the long term! If it doesn’t work now we will never build a system that is worth preserving into the future. Lets just do the easy stuff now and then migrate it if we ever need to. I believe this is my final word on persistence of GUIDs

Good point! But instead of questioning the trustworthiness of GUIDs in general we should start thinking on ways to ensure or at least enhance it. So, how can we REALLY trust? Surely not based on promises. But based on facts!

Therefore, the only way such service providers (also archives, databases etc.) could prove their trustworthiness would be a specific audit procedure and certificates.

There are some initiatives working on this. For example the German nestor group. They developed a criteria catalogue for long term archives largely based on the OAIS model. Another check list is the TRAC list here. And there is the European DRAMBORA project which provides a repository audit method based on risk assessment'.

The TDWG GUID group was focusing on technology, which was good to demonstrate a GUID system can work for biodiversity informatics. Now that we face these 'new' problems with GUIDs (business model + trustworthiness) TDWG should continue, revitalize the GUID group and care about these issues.

A good starting point would be to develop such an audit procedure for LSIDs, based on the models mentioned above. TDWG (or some other organisation) could offer these audits and LSID authorities could get certified. This would enable them to demonstrate their trustworthiness, thus the persistence of their GUIDs. Such a certificate could electronically be assigned and added for example as a link to the LSID metadata.
And why offering such audits for free? Maybe revenues from audit certificates could be one pillar of a future business model for LSIDs .. or TDWG?

What LSIDs are good for

Recently, I found David Shorthouses blog post on LSIDs (Life Science Identifiers) where he reports that apparently several LSIDs from different data sources exist for a distinct (spider) taxon. His intended use for LSIDs was to use them instead of the real taxon names and to "confidently link names with other sources of information such as information about the type specimens, gene sequences, synonyms, specimens etc". But for this purpose he concluded LSIDs are useless without a centralized identifier registry which ensures that a taxon name has only one LSID.

Well, replacing taxon names by unique identifiers is of course not the most obvious usage of LSIDs. LSIDs are only useful to persistently link to a electronic ressource and are not suitable to link to an abstract entity such as a taxon name. Instead, LSIDs link to a set of metadata, which contains the information which was considered to be useful by a LSID authority. Thus, LSIDs for taxon names link to nothing more than to the electronic resource (metadata, data sheet) which represents the LSID authority's concept of this taxon (btw.: a very interesting article on the whole complex of taxon names, identifiers, authorities, 'real taxonomists' and 'name users' is the Nature paper written by Nimis (2001): A tale from Bioutopia.).

LSIDs are however extremely useful for taxon names when they link to an electronic ressource which serves as authoritative record for this name. Especially LSIDs for newly assigned names which are officially registered by the ICZN in ZOOBANK can serve as citation for the name. This is similar to the usage of DOIs for scientific primary data. R.Pyle for example linked his newly assigned fish names to ZOOBANK records by LSIDs in his recently published paper. An impressive demonstration how useful LSIDs for taxon names can be.