- From: Tim Berners-Lee <timbl@w3.org>
- Date: Wed, 24 May 2000 13:00:35 -0400
- To: <xml-uri@w3.org>, "Tim Bray" <tbray@textuality.com>, "Simon St.Laurent" <simonstl@simonstl.com>
From: Tim Bray <tbray@textuality.com> >At 05:13 AM 5/23/00 -0400, Tim Berners-Lee wrote: > >>>I am growing increasingly short-tempered at some of the shabby rhetoric >>>being used in this debate >> >>If this general complaint was prompted at all by my Goedel-Escher Bach > >Nope. I was using "rhetoric" in the formal sense - prose designed to >convince the reader/listener of some point of view - and complaining about >rhetorical techniques which claim not only that point of view X is wrong, >but that belief in X is evidence that its proponents are against URIs, >against relative URIs, against RDF, etc etc; all allegations that have been >made explicitly and in some cases repeatedly. Maybe I'm losing perspective >on this; all I can do is report empirically that this tactic is severely >irritating and not helpful in looking for the solution to a hard problem. I am sorry if that has seemed to be the -- am perfectly happy to have my arguments dismantled with reason. I have felt exactly the same when I have had my insistence that dereferencing namespace URIs not be prohibited be represented as an insistence that *everyone* dereference the URI! I know how it feels. I didn't think I had represented you as being against URIs or RDF. I understand that you do thing relative URIs are dangerous. We agree on most of this. I feel that the situation in which they are allowed AND are compared as strings is dangerous. (Something which you compare as a string is not a relative URI.) You can obviously get out of this by disallowing either behavior. We disagree about which we would prefer. We are in the awkward situation that we have all arrived in a situation that the URI spec and the Namespace space are basically in conflict, which no one likes. We need to get out of that situation. Simon St Laurent said, >I don't think GEB is what irritated Tim; I think it's more this: > >>I'm sorry, but if the spec (and attendant expert interpretation) >>disallows dereferencing of the URI, then prohibiting changing the spec >>on those grounds effectively prohibits dereferencing URIs. >>(was this an accident or a slick process play by an anti-URI group?) >(http://lists.w3.org/Archives/Public/xml-uri/2000May/0320.html) Yes, that was strong. I felt I had been just faced with a fait accompli for which no one was taking responsibility. 1. Many people insist (Megginson excluded) insist that it is OK to dereference a namespace URI - just not a goal of the NS spec. 2. The namespace comparison by literal string in the namespace spec leaves a serious technical inconsistency in systems which do both "at different levels", as demonstrated by example cases, so I see (1) as being in conflict with the Namespace spec; but 3. I am told that is prohibited to change the namespace spec on the grounds of dereferencing requirements because that is not a goal specifically of the namespace spec Catch22. I was frustrated. If the consensus were "dereferencing namespace names is bad" a la Meggison, Keshlam, then there is one argument, against which one can argue. But the "but we are not prohibiting it anyway" line doesn't go down well when you can't build a technically sound system on top of the spec and comments on that basis are deemed unacceptable. So, while I would not brand you, Tim, as anti-URI, somehow we have to get a solution which is consistent with the URI spec. The namespace spec, we now realize, breaks relative URIs as defined in an IETF specification of long standing and great use. The fact that an XML group has met "consensus" that relative URIs should be allowed and compared as strings simply breaks the concept of relative URIs. It is basically an invalid reference the URI spec. You don't want Dan Connolly to talk about web architecture, but I am sure that if you found an XML application actually abusing the intent of XML you would feel there was a problem. Tim Bray and I agree on most things - about his first 3 points. We disagree about relative URIs. I think the question for me rests on whether there really are any examples of documents which would break if we changed the namespace spec. If there are, then I guess we have to warn that relative URIs are interpreted inconsistently and should be avoided. Simon again: >or Dan Connolly more-or-less continuous but vague discussions of 'web >architecture'. Please, there is a specific problem. We have an RDF specification and an XPath specificatoin which have interpreted the spec one way, and a lot of people implying that that was wrong. The fact that the namespace was a web Resource was a general assumption which had been made about how things fit together. That you can call architecture. Dan tends to be the first one to give you a very precise failure case to illustrate his point. >It's not been a pretty discussion, to say the least. This discussion has unearthed quite a lot of misunderstandings. There are probably still more. If rhetoric has been unfair, I would not say anyone has had a monpoly on it. There is a strong flavor of "wev'e discussed this before and you are wrong" which does not rank with me as well as sound arguement. But let us let that wash under the bridge and sort out the solution. >Simon St.Laurent Tim BL
Received on Wednesday, 24 May 2000 13:18:53 UTC