- From: Tim Berners-Lee <timbl@w3.org>
- Date: Tue, 30 Oct 2001 08:12:01 -0500
- To: "Roy T. Fielding" <fielding@ebuilt.com>, "Harald Alvestrand" <harald@alvestrand.no>
- Cc: "Al Gilman" <asgilman@iamdigex.net>, "Eastlake III Donald-LDE008" <Donald.Eastlake@motorola.com>, "'Rob Lanphier'" <robla@real.com>, <uri@w3.org>, "Dan Connolly" <connolly@w3.org>, "Larry Masinter" <LMM@acm.org>, "Dan Zigmond" <djz@corp.webtv.net>, "Rich Petke" <rpetke@wcom.net>
----- Original Message ----- From: "Roy T. Fielding" <fielding@ebuilt.com> To: "Harald Alvestrand" <harald@alvestrand.no> Cc: "Al Gilman" <asgilman@iamdigex.net>; "Eastlake III Donald-LDE008" <Donald.Eastlake@motorola.com>; "'Rob Lanphier'" <robla@real.com>; <uri@w3.org>; "Dan Connolly" <connolly@w3.org>; "Tim Berners-Lee" <timbl@w3.org>; "Larry Masinter" <LMM@acm.org>; "Dan Zigmond" <djz@corp.webtv.net>; "Rich Petke" <rpetke@wcom.net> Sent: Wednesday, September 26, 2001 2:05 AM Subject: Re: Excess URI schemes considered harmful > > roy: the worst thing with relative URIs is that at any time, there is only > > one base. If you have stuff from 2 naming trees at the same time, you're in > > trouble. That is what namespaces are for, of course - when you have a langauge with namespaces. > On the contrary, I can have a hundred different bases for which a given > relative URI can be resolved, just as we can have a hundred different > repositories for standard MIME types. Since I only use the relative > parts within the protocol syntax (unless it is a non-standard extension), > it really doesn't matter to me what the base is, provided that I pick one > (or allow the user to configure one) that points to at least one existing > namespace that is managed by the IETF. "text/plain" is a relative URI. > For this type of identifier, I simply don't allow relative names outside > of those within the standard namespace, and I pick the base according to > an algorithm that is different from web page retrieval. Having an http: URI doens't mean you *have* to retrieve a page every time the URI is encountered. There are all kinds of ways you can know what it stands for from having it programmed in to having a persistent cache (catalog). But it *does* provide a way of IANA providing *definitive* information about that URI. Such as binary grammars, and conversions between types, etc, [pointers to] definitive specs. Because when you use DNS to look up the URI owner's server and ask that server for a representation of the resource, what you get, inhaving been authorized by the owner of the URI, is to that extent authoritative. And that is useful. > Personally, though, I don't see any reason to standardize all IANA field > values as some sort of URI or another. A media type is simply an identifier > within a given context, and the Content-Type field name is more than > sufficient to establish that context. For every field which has a extensible set of string values, it is traditional to invoke IANA to deal with the problem. This doesn't scale. I see a purpose for IANA both historically and in the future as an accessible persistent store, and a link with the standards process which provides a set of specs which have a claim to quality and interoperability. But it should only be thought of as covering those resources which need to be common and shared globally. There should also be room in the architecture for ones which are local or experimental, etc. This is achieved by malking the IANA space a subset of URI space, a subset whose special place is special because of the way IANA is run. > ....Roy >
Received on Tuesday, 30 October 2001 08:12:09 UTC