- From: Matthew Dovey <matthew.dovey@las.ox.ac.uk>
- Date: Tue, 25 Mar 2003 22:01:23 -0000
- To: "Mike Taylor" <mike@indexdata.com>, <rden@loc.gov>
- Cc: <www-zig@w3.org>, <carrol.lunau@NLC-BNC.CA>, <slavko@mun.ca>
> I would like to know what the disadvantages _are_ before > agreeing to discard this approach! At first sight, it seems > admirable. From my recollection of the comspec approach - it was partly that we weren't really defining different record syntaxes (its always XML) so recordSyntax wasn't the right field. The fact we'd made that mistake with MARC was not seen as a good reason for repeating the error. There are some issues (not unsurmountable) - it could lead to a very long list of registered XML schema (I can think of a dozen bibliographic XML schemas off the top of my head). This could be tempered by not registering anything until someone needs it. Also how to handle private XML schema - using this OID prefix or a private one. Does that mean you have to become a registered Z39.50 implementor (and get a private OID prefix) just to use a private XML schema. Also, if we already have a persisent URI for a schema - why create yet another one! > Was I out of the room? :-) I had it reported to me since I wasn't at the particular ZIG (I think it was a ZIG or two ago, not the last one, but I may be wrong). > The element-set name really doesn't seem like the right place > for this at all, but I admit to not having any real > recollection of that discussion, so I may well be forgetting > some compelling point. I agree, but I thought comspec fitted the bill (I think I was one of the proposers of that). But there is v2 to still support apparently... > > I understand the weakness of this approach, it won't scale > well if we > > have thousands of schemas, but at the moment we don't, and > we need a > > simple and quick solution. > > I don't see a scaling problem with this at all I can't see a scaling problem either. > (The only problem is that, since the DCMI > steadfastly refuse to define an XML Schema for Dublin Core, a > dozen different projects are busily at work defining their > own, subtly incompatible, versions.) Exactly - that's were having to register OIDs for each one (rather than just use their URI) may not work! Matthew
Received on Tuesday, 25 March 2003 17:01:31 UTC