- From: Paul Prescod <paul@prescod.net>
- Date: Thu, 10 Oct 2002 14:12:23 -0700
- To: David Orchard <dorchard@bea.com>
- CC: www-tag@w3.org
David Orchard wrote: >... > > But won't the authors of the the http namespaces now want client software to > do something with the REC-RDDL file? Presumably there is some new kind of application that is enabled by the RDDL deployment. For instance it becomes possible for XML editors to give "friendly names" of elements or auto-discover datatypes. If your client app needs this new feature then it tries to dereference URIs it sees. If it doesn't, it just does string comparisons as today. >... > This seems like a most succinct summary. And it still causes me confusion. > > Given the lack of constantness, how does a user of a thing know when it > changes from one form to another? Answer #1: I would say that either GET returns 404 or it doesn't. If the provider of the URI knows whether GET should return useful information or not, then they could add that as metadata surrounding the link. Answer #2: If GET always returns something (either because we make that a Best Practice or because we define "404" as "something") then there is no confusion. It's just a question of whether GET returns something interesting to a particular client or not. This comes back to either testing or using surrounding metadata. > ... Presumably we want the software to do > different behaviour depending upon whether it's concrete or abstract. I disagree. Either you work with the thing "by identity" (string comparison) or you need to know something about its state that can only be answered by dereferencing it. I don't think that a single application (or application feature) needs to shift from one mode to another. If it works with the former it should use the former. If it needs the latter it should use the latter. If it COULD USE the latter but fallback to the former, it probes and does the right thing. > ... And > the only tool we have is the name. And surrounding markup/metadata/assertions. Paul Prescod
Received on Thursday, 10 October 2002 17:13:12 UTC