During the two GUID workshops there were really two reasons why DOIs seemed unworkable on a per-record basis. One was the cost. The other was the requirement for the central DOI administrator to have knowledge of every single DOI so it could be handled separately. (This seemed a ridiculous overhead for the expected use case of assigning identifiers to records as they are added to a database.)

If the prefix-based model is indeed available, it could provide most of the expected benefits we hope to achieve through LSIDs. However I suspect that we would end up having to do almost everything to support these DOIs that we have to do for LSIDs. We would still need local resolvers to return the record corresponding to the post-prefix identifier, and we would still need to wrap the DOIs with an HTTP resolver to use them appropriately in the semantic web. The big gain could be a central registry to handle relocating of services.

Donald

Donald Hobern, Director, Atlas of Living Australia CSIRO Entomology, GPO Box 1700, Canberra, ACT 2601 Phone: (02) 62464352 Mobile: 0437990208 Email: Donald.Hobern@csiro.au

-----Original Message----- Date: Tue, 25 Nov 2008 11:59:06 +0100 From: "Markus D?ring (GBIF)" mdoering@gbif.org Subject: Re: [tdwg-tag] Are DOIs expensive? To: "Gregor Hagedorn" g.m.hagedorn@gmail.com Cc: Technical Architecture Group mailing list tdwg-tag@lists.tdwg.org Message-ID: F245595E-8FF9-4277-AE01-88AD4B3F9708@gbif.org Content-Type: text/plain; charset="windows-1252"

That is very interesting, Gregor. Apparently it is up to the registration agency on how much they charge (and not the central DOI). There are currently 8 agencies for different areas with the German TIB issuing DOI prefixes for "registration of scientific primary and secondary data" at the cost of 250? per prefix (which allows you to create as many DOIs as you like).

http://www.doi.org/handbook_2000/registration_agencies.html#8.2 http://www.doi.org/registration_agencies.html

It would also be possible to setup a new registration agency for biodiversity data if we feel this is more suited to our needs. There are several costs associated with this though, an annual membership fee ($35.000), a "franchise fee" for each newly registered name ($0.04/ doi) and a maintenance fee ($0.005/doi). So this is clearly much more expensive and would be around 9 million dollars each year for 200 million occurrence records in GBIF. http://www.doi.org/handbook_2000/registration_agencies.html#8.8

So I guess this is what was investigated before and which is far too expensive. But it should be worth consulting the TIB registry, 250? per publisher doesnt sound bad at all.

Markus

PS: I am expressing my personal thoughts in this conversation and not GBIFs official policy.