IVOA Identifiers Working Draft

[Tony Linde]

Hi Will,

I know what you mean but that approach will plunge us into the morass of
'when is a copy a copy?'. If it is the same bit for bit, yes. What if the
data is the same but it is implemented in a different dbms? What if the
master has changed but the change has not been replicated? What if the copy
has been changed but not in the area I want to query? And what constitutes a
change: a bit, a record, a block of records? Etc. 

All these issues are best addressed in the metadata of the resource registry
entry. Separate resources should have unique IDs and we can address 'copy'
issues with metadata fields like 'originalName' or 'copyOfResourceID' or
whatever.

I *do* think the issue of how we identify and classify resource copies
(*and* other relationships) does need addressing, but perhaps not yet. 

I personally think the new registry schema that Ray and I were charged with
working up into a proposal (and will be out shortly) and coming up with
standard harvesting protocols are more necessary to get a working
international VO demo built by the end of the year.

Cheers,
Tony. 

> -----Original Message-----
> From: Wil O'Mullane [mailto:womullan at skysrv.pha.jhu.edu] 
> Sent: 04 September 2003 17:35
> To: Tony Linde
> Cc: 'Gretchen Greene'; 'Ray Plante'; registry at ivoa.net
> Subject: Re: IVOA Identifiers Working Draft
> 
> 
> > 
> > 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.
> > 
> I was thinknig of another model where the identifier was not 
> tied to a physical entity. We intorduce a lookup service 
> which returns physical identifiers/URIs for given 
> Identifiers. The lookup for an id could return multiple 
> physical represetations .
> 
> The service url otherwise is a unique identifier ...
> 
> wil
>