Registry discussion

[Ray Plante]

Hey Roy,

On Mon, 15 Sep 2003, Roy Williams wrote:
> http://www.ivoa.net/twiki/bin/view/IVOA/IdAmendPtA
> This looks like reinventing the DNS system. A lot of extra work so that
> people can choose what their IDs look like. I think the ID should reflect
> where you can go to resolve it. If the ID says roy.org, then there should be
> a resolver service at roy.org on the internet. There may also be other
> places that have harvested roy.org, eg ivoa.net or us-vo.org.

Yes, it is true, I think, that this is a reinvention of DNS.  This is in 
direct response to the desire of our collaborators that it should not be 
necessary that a naming authority be required to run a registry service to 
create identifiers with its authority IDs.  I agree that your idea 
simplifies the distributed bookkeeping; however, to provide greater
flexibility, more infrastructure is needed.  

Note that it will be easier for people to create new AuthorityIDs than DNS 
names; they need only register the AuthorityID with a global registry.

> http://www.ivoa.net/twiki/bin/view/IVOA/IdAmendPtB
> Whenever I see ivo:// at the start of an identifier, I will assume some part
> of the ID corresponds to an item in a VO registry. If someone wants
> identifiers which are not registered, please can they use a different
> format.

A repeat of what others have said:
  *  data providers may wish to start tagging resources, particularly 
     datasets, with RM metadata including an Identifier with an authID 
     they control, but not register it right away necessarily.  It could, 
     however, be exposed through some service (e.g. SIA).  If the provider 
     uses a local identifier format, then they will have to change it 
     when they do decide to register it.  

  *  assignment of identifiers should be independent of registration.

> http://www.ivoa.net/twiki/bin/view/IVOA/IdAmendPtC
> I don't see why we need to be concerned with persistence and location
> independence at this time. Looks like everything in this document can be
> safely left to next year once the registries are in use (once we know what
> we are doing).

By addressing persistence and location now, we have a straightforward way 
to provide compatibility with the ADEC DataResolver work (which you agree 
is a top priority).  Part D recommends that data providers based their 
ADEC dataset identifiers on LogicalIdentifiers.  

cheers,
Ray