- From: Williams, Stuart <skw@hplb.hpl.hp.com>
- Date: Thu, 15 Aug 2002 18:10:34 +0100
- To: "'Dan Connolly'" <connolly@w3.org>
- Cc: "'Ian B. Jacobs'" <ij@w3.org>, www-tag@w3.org
Hi Dan, > -----Original Message----- > From: Dan Connolly [mailto:connolly@w3.org] > Sent: 14 August 2002 19:52 > To: Stuart Williams > Cc: 'Ian B. Jacobs'; www-tag@w3.org > Subject: RE: 13 Aug Arch Doc available for review > > > On Wed, 2002-08-14 at 13:07, Williams, Stuart wrote: > [...] > > URI and URI References > > ---------------------- > > > > [skw-2002-08-14-02] > > > > I am very wary of the attempt to redefine the terms URI at the end of > > section 1.1. I think the terms URI and URI reference have been a source of > > problems in the past and will likely be a source of problems to come. But > > IMO, the definition of TAG-URI = AbsoluteRFC2396URI+OptionalFragment will > > make this document almost unreadable to anyone with a deeply infixed notion > > of URI based on current conventional usage. > > I think I can count the number of people that know that > http://example#foo > isn't a URI on my fingers. OK, maybe I need my toes. So somewhere between 2^10 and 2^20 then... fine motor control over my toes is a little lacking ;-) [Oh and if the number is less than 9 the TAG certainly has a problem :-)] I think anyone who has made or is making a serious attempt to get their head around the architecture of the web (rather than simply use the web from a browser) has encountered these terms and has indeed tried (and probably to a large extent succeeded) in understanding the syntactic relationship between AbsoluteURI, RelativeURI and URIReferences... and to some extent the torture history of URL, URN and URI. > (Keep in mind most people have never even heard of a URI; > they hurl URLs around, or just talk about web > page addresses). So... what web page is addressed by: http://www.w3.org/2001/tag/2002/0813-archdoc#Intro ? Is it the same web page that is addressed by: http://www.w3.org/2001/tag/2002/0813-archdoc#identifiers ? > And I agree that the $Date: 2002/08/13 20:19:49 $ draft > doesn't help much: it assumes that you've got your > head around RFC2396, and then throws you a curve-ball. > > But there is, I think, a fairly simple and appealing > architectural view in which http://example#foo > is in the same class as http://example, but > ../foo is in a very different class from > http://example#foo . Does this rendition of it appeal to you? Well... maybe... its worth exploring a bit... but you've also introduced another troublesome word, 'class', which might solve a 'problem' or may just be a symptom of the balloon bulging in a different place. I'd also have to ask, when speaking of 'class' are we speaking of the 'class' of the identifiers or the things that the identifiers identify... so, http://example#foo, http://example and ../foo are all URI references, the first two share a common absolute URI and the third one is a relative URI. If we are speaking of the 'class' of the referenced thing... there seem to be two views: a) i) The 'class' of thing identified by an absolute or relative URI is a Resource (or choose another term of the same concept). ii) The 'class' of thing identified by a URI reference that contains a fragment id is not a Resource, and we have no established term for what that 'class' of thing is. Candidates for that class of thing have been, 'item', 'sub-resource', 'part' and even 'fragment'. b) i) The 'class' of thing identified by a URI Reference is a Resource. ii) Some characteristics of a Resource vary depending on whether or not the is an unescaped '#' character in the URI Reference used to identify the resource [haven't got this right... but there is something of this ilk.]. > introducing URIs [was: 13 Aug Arch Doc...] > Dan Connolly (Tue, Aug 13 2002) > http://lists.w3.org/Archives/Public/www-tag/2002Aug/0138.html Haven't got to this one yet in crunching the backlog from vacation... > [...] > > > [skw-2002-08-14-03] > > > > I am also very wary about asserting that the thing identified by > > URI+fragment is a resource (which is what happens once you accept the > > redefinition of URI). On one level I could live with it justified on the > > basis that anything with identity can be a resource (RFC 2396 says something > > like that). > > > > However, resources can have fragments, > > Be careful with your quantifiers. To be clear, yes, I think I was being careful in chosing the word 'can', but I agree with your elaboration below which makes the meanin of 'can' very precise. > > There exists resources R and F such that F is a fragment of R. Yes... this (above) *is* what I meant. > > but *not* > > For all resources R, there is a resource F such that > F is a fragment of R. > > let alone > > Given a URI I for a resource R, there exists a resource F > and a URI J, computed from I alone, such that I identifies F > and F is a fragment of R. > > > and if I have a resource identified by the URIRef > > http://example.com/someResource#otherResource, how do I > > reference a fragment of that resource (assuming it has one)? > > OK, assuming it has one, I can coin a new URI > > mid:2002-08-14.thismessage@w3.org#abc > (pretend that's the MID for this message) > > to refer to it. But that's tant amount to saying that to make it a Resource (so that you can refer to some further nested fragment within it) you *have* to give it assign it a URI. [I accept that you wouldn't have to do that for 'resources' were not fragmented further.] > > The URI syntax > > does not provide me a way to do that... > > Depending on how you meant the quantifiers in > your sentence to work, either (a) yes, it does, see above, > or (b) no, it doesn't, but who cares? Well if I know the identifier for R eg. http://example.com/someResource#otherResource and I know the fragment, F, of R that I want to reference, say 'someFragment' RFC 2396 does NOT give me a way to construct an identifier for the Resource F from the identifier for R. Or... maybe it does *iff* independently F is also a fragment of the resource, Q, identified by http://example.com/someResource. Then http://example.com/someResource#someFragment is an identifier for F. We do lose the notion of being able to parse out the 'containment' relationships beyond the top level - but maybe that is not important (cf. pwd - where am I now?). Hmmmm.... I need to chew on this some more. Maybe my comfort levels will improve :-) > > so maybe one heads off may be to map > > the name of this resource into some new URI scheme eg. > > resource:http://example.com/someResource?frag=otherResource so that it can > > now attract further fragment IDs. The particular design would need more care > > so that it would recurse nicely. But that feels like a horible kludge. > > > > Anyway... bottom-line. On the TAG telcon, we did discuss distinguish between > > the sort of thing that a URI references and the sort of thing that a > > URI+fragId references. I think there was consensus (although I'm not sure we > > really tested it) that need distinguish terms for the referent... > > I got the impression we were headed the other way. I probably lost track of the conversation... somewhere there was the notion of UII's (Universal Item Identifiers), initially with 'items' being the sort of thing referenced by URI+fragmentId. But there seemed to be some desire to call such things 'resources' and then a dangling question of what then to call the things referenced just by absolute or relative URI. I don't think we really closed on whether there were one or two concepts that needed names. The single concept case is easy... its a resource. The two concept case is harder because of a desire on the part of some to switch 'item' and 'resource' and then find a better term for 'item'. > > but I > > think it would be folly to use the term 'resource' to denote the sort of > > thing referenced by a URI+Frag since that is commonly held to be the sort of > > thing that is referenced by a URI. > > Please take a look at my "introducing URIs" message to see > if it really looks like folly. Next on my to do list... but that will be tomorrow (at the earliest). > -- > Dan Connolly, W3C http://www.w3.org/People/Connolly/ Regards Stuart
Received on Thursday, 15 August 2002 13:12:58 UTC