[MOBY-dev] rfcf 5. LSIDs and Service Instances

[Roman Roset]

 >> 5. LSIDs and Service Instances
 >>
 >> a..If a service instance is registered by 2 different
 >> authoritys, should they have the same LSID?
 >>

 > Only if we can assure they are the SAME service (replicated services),
 > which would require external help.

   If I am not wrong, among other things, the lsid of a service
would have to serve to obtain the RDF and  I think
that it should be interesting to have one-to-one relationship between 
RDF and  service (or instance of a replicate service).  For instance, if 
the RDF of a replicated service needs to save information about a 
concrete instance (such as the service is temporally unavailable, 
cpu's, os, etc...) These values could be different for two "same 
services".  (but maybe I don't have clear what
are same services - My understanding is that, two services are the same 
if  given the same input then the output of both is exactly the same).

   However, I have some doubts about the way to specify the LSID of the 
services in biomoby. There are an example in the biomoby metadata-resolver:

urn:lsid:biomoby.org:serviceinstance:www.illuminae.com,getSHound3DNeighboursFromGi

(The LSID syntax is urn:lsid:authority:namespace:identifier:revision)

Why is the authority "biomoby.org" and this one is not www.illuminae.com?
Why is the service name separated of the lsid?

If we use biomoby as authority but we wrote the services as identifiers 
then an approach to denote more of one instance (or replicated service) 
  into the same lsid could be using the revision field. For instance:

urn:lsid:biomoby.org:serviceinstance:getSHound3DNeighboursFromGi:1
urn:lsid:biomoby.org:serviceinstance:getSHound3DNeighboursFromGi:2

But sincerely,  my opinion is that it's better to use an lsid like:

urn:lsid:www.illuminae.com:biomoby.org:getSHound3DNeighboursFromGi
urn:lsid:www.other.com:biomoby.org:getSHound3DNeighboursFromGi

   With this kind of LSID the MetaData-resolver could know
that if the namespace is biomoby then the element to resolve is a 
service, and we should have one LSID per service.

 >> b. if the service instance is registered in 2 registries,
 >> should the LSIDs be the same?
 >>

 > No, they shouldn't, unless they are registered by the same authority.

Is the RDF of this sevice shared by both registries? Is the response is 
yes then I feel that the LSID should be the same.
(and the opposite, if the response is not the LSIDs should be different.)

 >> c. will modified service instances have versioning
 >> information?
 >>

 > I think it should, because versioning information could help in the use
 > of services from different registries, or the integration of those
 > registries.

 From my point of view, for one lsid of a service (independently of what 
thing is identifying exactly) there are two ways to understand the
versioning information.  I think that the version has to "explain"
(useful) changes in the meta-information.  But the service 
metainformation (the RDF) can change at two levels: the values of the 
predicates and the schema.  I think that versions should be useful (at 
least for the agent) in the second case.

-- 
Roman Roset Mayals
Technical Coordinator
   Instituto Nacional de Bioinformatica (www.inab.org)
Barcelona Supercomputing Center Node (www.bsc.org.es)
c/. Jordi Girona 1-3
Modul C6-E201                     Tel.  : 934 011 650
E-08034 Barcelona                 Fax   : 934 017 014
Catalunya (Spain)                 e-mail: rroset at lsi.upc.edu
Note: I'm on vacation until August 21th