- From: Ayhan ULUSOY <ayhan@drexel.edu>
- Date: Wed, 5 Jul 2000 23:42:44 +0200
- To: <xml-uri@w3.org>
Summary of problems with Literal, Forbid, and Absolutize : 1) Literal (without Absolutize) *either* breaks the URI concept *or* requires a detachement of namespace IDs from URIs altogether; 2) Forbid brings about a new "kind" of URI *and* it forbids something *altogether* which may be desirable *sometimes* (but often not); 3) Absolutize (as it is today) may be dangerous for the *usual* namespace use that begs for uniqueness *and* it may break some existing documents (though this is quite unlikely). Besides, absolutizing *against the usual Base* may not be something that desirable anyway ------------------------- The CURRENT PROPOSAL(unspecify away) seems to leave enough room to breathe and de-stall many without closing too many doors. Here is such a possible door (assuming that there are people out there who are still interested to discuss the longer term). It is a kind of "Absolutize" but only in a more generalized way. ------------------------ *** The METAPHYSICAL PREFERENCES : Being able to use relative URIs with namespaces : 1) would suite more with the "uniformeness" of URIs. 2) may be desirable (as suggested by several scenarios)....but not necessarily against the usual Base (in the XBase sense). I would even say "often not" against that usual base. *** The SUGGESTION that goes with it : Further generalize XBase to include an orthogonal concept for determining the BASE. That concept would be the "class" of object (whether it may be retrievable or not, concrete or abstract) that a given URI is supposed to identify. Namespaces would be such a "class", whose default base for any occurence in a given document COULD (but not necessarily) be seperately declared via this extension to XBase. This may be quite handy for stylesheets, out of line links and such too. These things need retrieval at one point and being able to have seperate base URIs -regardless of the container- may be quite nice. Well, of course, *having* to have a such seperate base is not nice at all. But this issue can be resolved. The syntax could be constructed such that XBase scheme does not need to know the existence of such and such class. The "concerned" spec, or application can just include that within, while also referring to XBase. That would mean the use of Namespaces to achieve the desired naivity of XBase vis-à-vis the declared classes. **** The PROBLEM with the SUGGESTION : One possible problem with this aproach is that it might require XML-namespaces and XBase to be heaviliy cross-dependent (if not the same spec).
Received on Wednesday, 5 July 2000 17:43:31 UTC