- From: Jonathan Rowell <bigrat18@hotmail.com>
- Date: Wed, 26 Mar 2003 13:40:42 +0100
- To: rden@loc.gov
- Cc: www-zig@w3.org
Hi all! If I may, as an outsider, be allowed to put in my penny's worth on this subject, since it interests me greatly. Ray wrote first: "That's my attempt to summarize the issue. My proposal is to adopt the approach used by SRW when faced with a similar problem. For the esn, use a URI that serves as an identifier for the desired schema (you don't have to call it a namespace)." and later to Diana Dietz : "if you try to go to: http://www.loc.gov/zing/srw/dcschema/v1.0/ you'll get "Not Found" because it's not intended as an actionable uri but as an identifier, the same way that an oid is an identifier. We *can* guarantee persistence of this uri, as well as un-ambiguity. If a new version is developed, there will be a new uri, maybe http://www.loc.gov/zing/srw/dcschema/v1.1/, but the old uri (http://www.loc.gov/zing/srw/dcschema/v1.0/) will continue to identify version 1.0." Now for me that's a namespace. If you want to validate a document using a DTD or a schema, you'd cache the thing locally (to avoid fetching it over the net) and match an identifier in order to use it. This is the namespace. Everybody does it, XSL, XSD, Marc21. The use of XML defines a language. This becomes fixed like the oids. But it's worse than that, the structure of the contents becomes fixed. Before one could cheat a bit by returning "something like". But not anymore. You have to ask the server to return the data in that XML language and either it does or it can't. That's it. Theo van Veen wrote : "The real problem is that XML schemas are not sufficient to express what you what to express. We had a similar discussion also at the ZNG list. There should be a way (a generic more schema) to request DC records without caring about the small differences in different versions." Exactly! We have this today. What do you think I'll get if I ask a Norwegian server for a German MAB record? (What do think I'll get if I ask a German server for the same!) As an implementor I'd bang over the identifier for the format, get the XML back, validate it syntatically, look at the namespace provided, look up in a table of namespaces for an XSL transform and transform the XML data into the required format and throw it at the user (poor swine). Ray wrote : "One suggestion is to assign object identifiers subordinate to 1.2.840.10003.5.109.10. This idea has a number of disadvantages and unless someone wants to pursue this approach I'd prefer to discard it." Seems to me the only way to stop a greater proliferation than you currently have is to allocate oids for all XML formats. The oid authority should insist that, in order to get an oid registered, there must be an authority which maintains a namespace and a DTD or schema to go with it. The latter the implementors may cache. The problem is the update. You ask for a particular format and get a different one. This in fact is no worse than today. The oid for German MAB actually gives you MAB2. Is this not the case for DANMARC and the rest of Scandanavia? You always get the new format. If however the namespace is mandatory in the data returned, we have a chance of sorting it out using very powerful tools like XSL. Jonathan >From: Ray Denenberg <rden@loc.gov> >To: zig <www-zig@w3.org> >CC: Lunau Carrol <carrol.lunau@NLC-BNC.CA>, slavko@mun.ca >Subject: requesting XML records >Date: Tue, 25 Mar 2003 14:38:12 -0500 > > >We need an agreement among Z39.50 implementors on >how to request xml records. This is, in >particular, an urgent issue for Bath profile >implementors. >......................... > >--Ray > _________________________________________________________________ Help STOP SPAM with the new MSN 8 and get 2 months FREE* http://join.msn.com/?page=features/junkmail
Received on Wednesday, 26 March 2003 07:40:49 UTC