Re: requesting XML records

Mike Taylor wrote:

> I have to say that I don't see this as a strong argument at all.  I'm
> sure that the inventors of HTTP would be amazed to see it now being
> used as the substrate for SOAP, and the history of computer science is
> littered with similar happy (or in some cases unhappy) accidents.  For
> that matter, I didn't envisage the Zthes profile being used to
> navigate phylogenetic hierarchies, but I'm glad people are doing it.

Sorry I don't buy this reasoning  -- you're comparing, on the one have,
where an application is invented and people find other useful thing to use
it for; with, on the other hand, an identifier, intended to identify
something specific (although I concede we don't know exactly what that is)
and we're talking about using it to identify something completely different.
(Sort of like, say I need a classification system for birds, so I use LC
numbers -- e.g.  the "Inyou California Towhee" gets assigned QL696.P2438R428
1998  -- which is the LC number for a book about that species.)


> > My suggestion is that we continue (as proposed) to think along the
> > lines of letting the element set name take on schema name values.
>
> Doesn't it?  That comes as a complete surprise to me.

Again, the out-of-the-room effect. We had this discussion in Dublin.



> Hmm.  Then we will need either (A) a central registry for schema-like
> XML element-set names, or (B) as with clasic Z39.50 practice, an
> understanding that all element-set names other than a tiny hardwired
> set ("B" and "F") are undefined except in the context of a profile.

Right, one or the other.

> I don't think the latter is what Theo wants at all, since (as I
> understand his requirement) he wants to cross-search targets that do
> not adhere to his profile, if he even has one.  (Right, Theo?)

Yes, I agree that's not what he want.

> So you do appear to be volunteering to maintain a register of these?

I'm suggesting that we find a way to keep the list of these manageable (and
if that's not possible, then use version 3/compspec) in which case registry
is not a major problem. But I'm not convinced that registration is needed or
is the right approach.

--Ray

Received on Friday, 28 March 2003 12:24:12 UTC