Re: requesting XML records

Mike Taylor wrote:

> I would like to know what the disadvantages _are_ before agreeing to
> discard this approach!  At first sight, it seems admirable.

1. It would treat these (DC, MODS, MarcXml, etc) as record syntaxes. They're
not syntaxes, they're schemas.  It was the Z39.50 community (not the w3c or
the xml community) who articulated this distinction, 10 years ago!  It would
be a big step backwards to now cast these as syntaxes.
2. oids aren't very user friendly.
3. oids aren't very widely used or implemented in protocols.


> > Yet another approach is to use the element set
> > name parameter to indicate the schema.  Actually,
> > this is pretty much what we agreed upon in
> > principle at the last ZIG meeting (nearly a year
> > ago).
>
> Was I out of the room?  :-)

Yes, about 3,000 miles "out of the room". (That'll teach you to miss ZIG
meetings!)  This was the Dublin (Ohio) meeting at OCLC last April.



> > But we didn't think this through carefully
> > enough.  Do we mean a schema, or a  namespace?
>
> (If we go this route at all) We _definitely_ mean schema.  Namespace
> is a complete red herring, and it is misleading even to mention it in
> this context.

I suppose I wasn't clear enough. I wasn't asking if we're *identifying* a
schema or namespace (the answer would be schema) but whether we're *pointing
to* a schema or namespace (in the sense of supplying an actionable url for
retrieval), in which case I'm suggesting that the answer is neither.
*Pointing* to a schema (i.e. with an actionable url) has the problems I've
pointed out in my response to Dana. Retrieving a namespace is meaningless.
So yes, we're *pointing* to a schema, but not with the schema's retrievable
url but with an abstract identifier (is my suggestion).


> I don't see a scaling problem with this at all -- in fact, I'd have
> thought scalability was one of the _strengths_ of this approach.

It's the same problem with oids -- you have to have a naming authority.  If
LC is the authority and we're talking, long term, about 100 schemas, no
problem. If we're talking about thousands then we have to distribute the
authority and that causes complication.  That's what I meant.

--Ray

Received on Tuesday, 25 March 2003 17:17:33 UTC