- From: Henry S. Thompson <ht@inf.ed.ac.uk>
- Date: Mon, 04 Sep 2006 12:43:54 +0100
- To: Bjoern Hoehrmann <derhoermi@gmx.net>
- Cc: www-xml-linking-comments@w3.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Bjoern Hoehrmann writes: > * Henry S. Thompson wrote: >>The terminology could, with hindsight, have been improved. What's >>meant is what the URI spec. [1] calls a _secondary_ resource. There >>is no requirement that secondary resources be in any straightforward >>way subparts of the primary resource identified by the pre-fragment >>part of a URI. > > That seems a creative interpretation of the specifications. A fragment > identifier identifies either one secondary resource or it is undefined > what it identifies; in contrast, the XPointer framework explicitly re- > fers to "one or more subresources". The conceptual model then also be- > comes rather silly; a pointer identifies a secondary resource; a pro- > cessor takes the pointer and a document and produces an identification > of a secondary resource. That's not how I read it -- for any URI scheme, documentation of a fragment identifier syntax has to explain _how_ it identifies a secondary resource. That's what the XPointer specs. do. The issue of discontinuous/multiple secondary resources is a separate matter, on which I still owe you a reply. >>I think this one is OK, in that the Framework is working from the fact >>that a single XPointer expression may contain _multiple_ pointer >>parts, to enable fallback. It's only an error if they _all_ fail to >>identify a subresource. Hence for a _single_ pointer part to fail >>softly and allow fallback, it should just not identify a subresource. >>And that's what the element() scheme prose you quote is doing. > > Yes, you can also read the text that way, that is the problem. > I note that this is further explained by > > For example, the following pointer part identifies the element > with an ID (as defined in XPointer Framework) of "intro": > > element(intro) > > If this pointer actually identifies said element, then the absence > of said element in the resource does not cause the pointer to cease > identifying a secondary resource, which would contradict what you > say above. This reminds me of tendentious claims that there's nothing wrong with the phrase "the current king of France is bald" -- it may have a 'meaning' in a formal-logic kind of way, but it's in practice useless (today). Just so, element(intro) doesn't identify anything if there's no such element -- the Framework is clear about that, as far as I can see. >>Not so -- since the SVG Viewer can correctly determine a secondary >>resource from the first pointer part, we're done. > > Then how do I know whether a pointer identifies a secondary resource? > The documentation of a scheme is not required to specify when a given > pointer part identifies a secondary resource, so scheme documentations > won't do that in the general case, and so there need to be other means > to determine whether they identify such a thing. Success/failure of syntactically OK pointer parts to identify is always an empirical question (just as for element(intro) above). The only way to find out if a pointer part identifies a secondary resource is to pass it to a implementation of the scheme and look at the result. > To give an example, if a #xpath2(...) pointer evaluates to an empty > node set, is the empty node set a secondary resource? I sure hope so, > but then it's a bit difficult to understand why element() identifies > no secondary resource just because the element referred to is absent. That question is surely up to the spec. of a scheme to make clear. Reasonable persons might differ on this one -- seems to me if I were speccing this one I'd treat an empty node set as failure to identify a secondary resource. >>Requiring "adequate specification" before registering a scheme amounts >>to insisting that scheme documentation go through the W3C process, >>which contradicts the purpose of the Registry, namely to allow >>XPointer scheme registration from _outside_ the W3C. The only basis >>on which the W3C went ahead with launching the Registry was that it >>required a bare minimum of Team resourcing. Any kind of QA >>requirement would require substantial resourcing. If W3C Members >>request this, it can and should go in to the pot to compete with all >>the other requests for Team resources, but that hasn't happened so >>far as I'm aware. > > Now that argument is just silly. You could simply require that the > description of a scheme is part of some standard published by a > recognized standards organisation, that would substantially raise > the amount of quality assurance at no, or in fact less, cost for the > Team. Where proprietary schemes are desired, and use of namespaces > is not, you could easily make a vnd.name naming scheme and reserve > names without a . for standards. Quite frankly, the flaws of the > xpath2 scheme, and in fact a number of others, are obvious at a > first reading, if reading proposed registrations is asking too much, > the whole 'review' part of the process is a farce. I'll take this suggestion to the Architecture Domain in the Team (note that the Registry is not the responsibility of the XML Core WG), but I have to say it's not a direction I personally would like to see us go - -- at the time XPointer went to REC we were asked, by several Members, to provide a way to broker contention for scheme names for public use. The current Registry provides just that. I understand that you would like to see it provide more than that, but that involves taking on responsibility, one way or another, for the quality of the schemes themselves, and I don't think that makes sense from a cost/benefit perspective. ht - -- Henry S. Thompson, HCRC Language Technology Group, University of Edinburgh Half-time member of W3C Team 2 Buccleuch Place, Edinburgh EH8 9LW, SCOTLAND -- (44) 131 650-4440 Fax: (44) 131 650-4587, e-mail: ht@inf.ed.ac.uk URL: http://www.ltg.ed.ac.uk/~ht/ [mail really from me _always_ has this .sig -- mail without it is forged spam] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.6 (GNU/Linux) iD8DBQFE/BF6kjnJixAXWBoRAnVjAJsFkdX2GXMh8qn6d/HbM3K3DcMbQACdEC5F /itOUzSADbCDvVXUwLYoI2A= =1PJ8 -----END PGP SIGNATURE-----
Received on Monday, 4 September 2006 11:44:05 UTC