Kevin,
Thanks a lot for filling in the details for the Index Fungorum LSID resolver prototype.
Interestingly enough, there is another GUID group (http://wiki.gbif.org/guidwiki/wikka.php?wakka=LSIDResolverNamespaces - coordinated by Donald) which task is to come up with guidelines for selection of LSID namespaces. That group will untimately come up with the answers to your question. However, since they just got started, they don't have any guidelines for us yet. Actually, this thread is likely to provide that group with use cases and scenarios for them to work on.
Anyway, in the absense of any namespace assignment guidelines, the LSID spec states that ids are supposed to be opaque and the namespace part is there just as a convenience feature for authorities to partition their internal databases. So, strictly speaking, we are not allowed to add any global semantics to the namespace. That means that is up to you (or IndexFungorum) to define what's in the namespace of the LSIDs you assign. That doesn't mean that we cannot come up with guidelines on how to best partition your internal database, though. But that's another story.
Assuming that Index Fungorum has complete autonomy over Fungi names (I assume it has as a nomenclator), and that it has the required software development and system administration skills, I think that IF would most benefit from a completely distributed LSID assignment and resolution schema. This implies that:
1) The authority identifier (and thus the DNS record) should be owned and set up by IF (i.e. indexfungorum.org, lsidauthority.indexfungorum.org, and lsid.indexfungorum.org are good choices while guid.org and indexfungorum.guid.org are bad options). 2) It's up to IF to define its namespaces and no external client (or no one at all) should rely on their choices of namespaces.
Any centralization in this case would yield in unnecessary delegation of tasks from IF to some external entities (whoever owns guid.org domain for example).
However, the community MUST agree on the format of the metadata returned by any Taxon Names resolver. That's exactly what the TAG (TDWG Architecture Group) is doing now.
Cheers,
Ricardo
Kevin Richards wrote:
I have briefly updated the Wiki for the IndexFungorum LSID prototype at http://wiki.gbif.org/guidwiki/wikka.php?wakka=LSIDResolverForTaxonNamesIF
As I was filling in the specs, it made me think it would be good to have a "suggested" list of LSID namespaces that different LSID authorities could use if suitable. At present the IndexFungorum namespace is "Names" - which may not be ideal. This will differ however for central LSID authorities and distributed authorities. Eg if a central authority "guid.org" is setup the namespace "indexfungorum/names" may be suitable for taxon name objects [a link to a glossary would be helpful here to explain what I mean by a name object], but for the distributed authority "indexfungorum.org" then just "names" may do the job.
Another alternative is to put the "distributed" authority in the authority name, eg "indexfungorum.guid.org" leaving it up to DNS to locate the authority, then leave the namespace simple, in this example "names", eg "urn:lsid:indexfungorum.guid.org:names:123456".
Any thoughts?
Kevin