IVOA Identifiers Working Draft

[Tony Linde]

Hi Gretchen,

> To sum this up,  there is no central control for id 
> assignment for VO resources?

That's right. The uniqueness is handled through the authorityID portion of
the key. The registry has the ability to ensure uniqueness within an
authorityID since it is the only place that a key containing that
authorityID can be created.

But how do we make sure no registry sets up an authorityID already managed
by another? This is something we'll need to address, possibly as part of the
discussions on reigstry harvesting.

One way might be that the metadata for a registry resource includes all the
authorityID's that that registry 'owns' - any new request for an authorityID
to any registry would only need to check against the list of authorityIDs
for all registry-type resources. We'd still need to handle the case where
the same authorityID is requested for separate registries before the
harvesting run of the second - I guess the second will need to disable
resources with that authorityID until the owner changes it???


Referring back to comments on mirrored resources, I think the only way to
proceed is that every resource in the VO is uniquely identified and whether
or not it is a mirror should be part of the metadata for that resource.

Cheers,
Tony.

> -----Original Message-----
> From: owner-registry at eso.org [mailto:owner-registry at eso.org] 
> On Behalf Of Gretchen Greene
> Sent: 05 September 2003 01:02
> To: 'Ray Plante'
> Cc: registry at ivoa.net
> Subject: RE: IVOA Identifiers Working Draft
> 
> 
> 
> Thanks Ray for your prompt reply.  Well yes,  you mostly 
> answered my question and it sounds fine what you have 
> suggested.  If I understand correctly then, this suggests 
> registries can serve resources (some mirrored of partially 
> replicated) with independent authority ids from other registries.
> 
> I guess for large archive centers,  the id's will evolve 
> consistent with the management of the registry and the main 
> goal up to this point is to establish the 'standard' form.
> 
> Are there any specific sw dev plans for networking registries 
> together across oceans and performing id matching/checking 
> [naturally it would be a key element for comparing] to 
> determine resource uniqueness?  Perhaps this is discussed at 
> the meeting.
>  
> To sum this up,  there is no central control for id 
> assignment for VO resources?
> 
> -Gretchen
> 
> 
> 
> 
> -----Original Message-----
> From: owner-registry at eso.org [mailto:owner-registry at eso.org] 
> On Behalf Of Ray Plante
> Sent: Wednesday, September 03, 2003 8:03 PM
> To: Gretchen Greene
> Cc: registry at ivoa.net
> Subject: RE: IVOA Identifiers Working Draft
> 
> Hey Gretchen,
> 
> On Wed, 3 Sep 2003, Gretchen Greene wrote:
> > This looks great to me,  I'm a little confused though 
> because in the 
> > last [NVO] metadata telecon,  I thought Arnold made a point that
> location
> > should not be a part of the id,  at least he said he 
> disagreed with me 
> > on that.
> 
> At the moment, whether IDs are location-dependent is a matter 
> of debate.
> 
> The WD represents a particular state of the debate (from 
> early summer).
> 
> We'll need to get concensus on this question before the WD graduates 
> to the next level.  
> 
> Personally, I think we need both location-dependent and 
> location-independent identifiers.  The former allows us to 
> distinguish between different replicas.  The current WD 
> basically punts on the latter, 
> calling it out-of-scope.  To me, the question is do we address the 
> location-independent version in this spec.  
> 
> > The question I have is that for people building registries,  for
> example
> > in our case...we registered several resources and assigned 
> 'authority 
> > ids' based on Bob's knowledge.  Now this is a special case 
> I know for 
> > our prototype/demo work,  but in the future...perhaps registry
> builders
> > decide (and I'm thinking of places like CDS that host LOTS of 
> > resources),  to swallow up resources.  HOW does the id assignment 
> > process work in this case?  The resource 'owners' are not 
> the same as 
> > the registry builders.
> 
> Perhaps we could compare a few scenarios:
> 
>   *  At NCSA, we have a publishing registry, where the user 
> picks their 
>      own ID, including the authority ID.  To register 
> resource, they first 
>      have to register their organization.  This is the 
> organization we 
>      consider as controlling the authority ID they have chosen.  
> 
>   *  At STScI, Bob assigned some IDs, including the authority 
> IDs.  In 
>      this sense, STScI controls the authority ID.  This is fine and 
>      consistant as well, even though it doesn't maintain the 
> resources.
> 
> 
>   *  At CDS, they archive and manage all tables published in 
> the journals.  
>      Others may maintain the copies of the same tables.  Given the 
>      number of resources they manage, they would likely set 
> up their own
> 
>      publishing registry.  They would create and control 
> their own IDs for 
>      those resources.  Others that have copies of the same 
> catalogs will
> 
>      register their own copies using their own authority ID, 
> different 
>      from what CDS uses. 
> 
> Does this address your question?
>  
> > Maybe there aren't that many resources for this to matter though?
> 
> (I think there are.)
> 
> cheers,
> Ray
> 
> 
>