Hi Bob,
Personally, I would hold on to pure LSIDs and avoid using the http proxy as a GUID as much as possible, just to make sure my clients are robust enough. However, the semantics of the proxy software and the RDF returned when the identifier is resolved allow clients to assume the http proxy is in fact a GUID.
Below is an example of RDF that you get back when you resolve the proxy LSID.
rdf:RDF <rdf:Description rdf:about="urn:lsid:ubio.org:namebank:11815"> dc:identifierurn:lsid:ubio.org:namebank:11815</dc:identifier> <owl:sameAs rdf:resource="http://lsid.tdwg.org/urn:lsid:ubio.org:namebank:11815" /> ...
In the case above, the LSID in its pure for is used as an identifier in *rdf:about* and *dc:identifier* (the namespace dc stands for Dublin Core.) However, the *owl:sameAs* statement asserts that both the LSID and its http proxy are equivalent and interchangeable. The proxy (the software at http://lsid.tdwg.org, for example) is the agent that actually implements that equivalence between the identifiers, without however asserting anything about resolvability of the proxied LSID. In my opinion, you can only assume that the proxy software will be able to resolve a given identifier as long as the proxy software itself and the LSID authority are operational. The proxy software cannot be deemed as an authority of that LSID. I think that this was your original concern, wasn't it?
In summary, there is no problem assuming that a proxy version of an LSID is a GUID. But there are operational implications of using either form. It will all depend on whether the client is aware of the LSID spec and its particular implementation. For example, if your client isn't aware of LSID spec, then you have no choice other than storing and using the proxy version to get back to the object associated with the ID. The client may use the pure LSID as a URI for internal calculations, such as graph merging and querying, but it won't be able to resolve it directly.
On the other hand, if your client is aware of the LSID spec, you have more options. You know how to get from an LSID to a proxy version and back. You can use it in your advantage, or not. You may want to store both versions so that your client is more robust to link rot, and so on.
Cheers,
Ricardo
Bob Morris wrote:
Is the http proxy a GUID?