Resource Identifiers: discussion synthesis

[Tony Linde]

Hi Bob,

I guess it depends on what you want to achieve with 0.8 of RSM. As I've
said previously, I don't really agree on the name of the document but
that may well be the least of the issues. Ray and I have converged on
some agreement over the ResourceID and I believe he is going to write it
up. There's still discussion on whether the registry as we're looking at
it now contains every resource or whether we separate out the
'supertypes' of resource into different registries. Then there's the
issue of the metadata formats or extensions, how they are defined and
how expressed.

If v0.8 is to encompass all of this then it'll be some time before it
can be ready (look how long it took for Ray, Roy and I to agree on the
ID - who knows what everyone else will think of the proposal?). 

I guess we need to solicit opinions first on the supertypes and
MFs/extensions ideas.

Cheers,
Tony. 

> -----Original Message-----
> From: Robert Hanisch [mailto:hanisch at stsci.edu] 
> Sent: 05 June 2003 21:39
> To: registry at ivoa.net
> Subject: Re: Resource Identifiers: discussion synthesis
> 
> 
> Ray, Tony, et al. -- I guess closure is being reached here, 
> though it still is a bit hard to tell.  I would like to try 
> to finish up v0.8 of RSM, but cannot really do this until we 
> have enough of an agreement on identifiers so that RSM is, if 
> not complete in its specification, at least not wrong.
> 
> We currently have three identifiers in RSM:
> 
>     Identifier (of the resource)
>     PublisherID
>     ServiceStandardID
> 
> All are currently described as URIs.  URIs are typically 
> based on an organization's IP domain, a string that is 
> guaranteed to be unique as a result of going through the IP 
> domain registration process.  So, isn't this good enough to 
> start with?  I think concerns about people registering 
> resources with non-unique identifiers are a bit spurious; I 
> mean, why would anyone do this?  The point of an identifier 
> is to help locate the resource. Using the URI for some other 
> resource or organization would have no benefit. People could 
> be malicious, I suppose, but to what gain?  It would seem 
> that DOS attacks on the VO would hardly be worthwhile.
> 
> At this point it is less important to me whether we stick 
> with a URI or use an arbitrary string provided by an 
> indentifier-issuing service.  I prefer the URI and I prefer 
> leaving the choice of the identifiers with the data provider. 
>  We can add an indentifier validator (are my chosen indentifiers
> unique?) or something as time permits.
> 
> I suppose I can modify RSM to read simply "Identifier: A 
> unique label associated with the resource" with similar 
> wording for PublisherID and ServiceStandardID, and refer to 
> the TDB Identifier Specification document.
> 
> As my 11-year old is fond of saying, "whatever."  We need to 
> move ahead with some agreement, even if it is an interim one, 
> and in any case, I hope it is a simple one.
> 
> Cheers,
> Bob
> 
>