- From: Sandro Hawke <sandro@w3.org>
- Date: Wed, 30 Oct 2002 08:25:13 -0500
- To: Dan Connolly <connolly@w3.org>
- cc: Patrik Fältström <paf@cisco.com>, www-archive@w3.org, Ned Freed <ned.freed@mrochek.com>, timothy@hpl.hp.com
Dan Connolly writes: > On Tue, 2002-10-29 at 13:35, Patrik Fältström wrote: > > > > The IESG is to publish draft-kindberg-tag-uri as an informational RFC > > shortly. Comments on the draft is to be sent to me personally and the > > authors. > > As I wrote to Sandro earlier... > > I certainly don't agree with: > > "But there are > drawbacks to URLs-as-identifiers: > > 1) Software might try to dereference a URL-as-identifier, even though > there is no resource at the "location"." > > (a) there's always a resource there; there might not be a representation > available. But (b) if there isn't a representation available, there > should be. A year and a half ago, when we were drafting this, Tim Kindberg phrased that text. The phrasing grated a little on my W3C-trained ear at the time, but I thought it spoke clearly enough to the wider audience. Let me try another phrasing: 1) With many URI schemes (such as HTTP) there is no indication in the URI itself whether web content is available representing the identified thing; the availability of such content must be determined via metadata or experimentation. While this flexibility can be very useful in some applications, in others designers may know in advance that for certain URIs such a representation will never be available. When this is known, use of a tag URI avoids the need to communicate extra metadata or perform the web-retreival experiment. Of course the text itself is really just informative, and could be dropped with no change to the specification itself. > i.e. names/identifiers take on meaning by use in protocols. > I don't see much value for making up names without some > sense of how they'll take on meaning. By "take on" meaning, I assume you mean in the social sense of establishing a meaning which is shared by various entities (as opposed to simply having a meaning defined in private by a creator). You're right that tag URIs have no such associated mechanism. We imagined that in some cases, separation of these mechanisms could be advantageous. My imagined application was the Semantic Web with floating semantics: the consensus meaning of an identifier is determined by considering various documents using it, along with document metadata. This approach mirrors natural language; Tim Berners-Lee recently said it should work in about fifty years; Pat Hayes said ten. In proposing tags, I'm not saying when it might work, just that I'd like people to be able to experiment as soon as they want to. Our current work (you and I at W3C) is aimed at applications where there *is* representative content available for retreival, and tags do not support that approach. Instead they support an alternative approach which can co-exist peicefully. In fact, if people want to not publish initial authoritative semantics, then use of a tag URI makes this clear; it means a system calculating the closure under inclusion-of-origin-documents doesn't need to do a bunch of retreival experiments. (In RDF one could provide metadata saying no such experiment will succeed, but that gives authors extra error-prone work. And what if the term creator wants to be sure no data is ever taken as creator-authoritative? Since triples can be freely dropped, that information cannot be maintained.) > And I don't see why the existing URI schemes don't work > just fine for naming stuff. There are people arguing that it is inappropriate to use each of the other approaches in some circumstances. Can you use an HTTP URI to identify a dog? Some say no. Can you use an HTTP URI-Reference to identify a dog? Some say the dependence on media-type information makes that also a bad idea; the naming is no longer uniform. Tags at least get one clear of the controversy of retargetting existing schemes. They also get one clear of all the useful machinery of the web, which is too bad. I'm not saying which way we should go; I'm just saying we should have tags so we can properly explore the territory. > FYI, the W3C TAG has been doing work in this area; it's > not finished, but a stake in the ground is: > > "Describe resources: Owners of important resources (for example, > Internet protocol parameters) SHOULD make available representations that > describe the nature and purpose of those resources." > -- http://www.w3.org/TR/2002/WD-webarch-20020830/ > > > Regards, Patrik > > Area Director, Applications Area, IETF > -- > Dan Connolly, W3C http://www.w3.org/People/Connolly/ Sandro Hawke, W3C http://www.w3.org/People/Sandro/
Received on Wednesday, 30 October 2002 08:26:42 UTC