W3C home > Mailing lists > Public > xml-uri@w3.org > May 2000

layering issues (was which layer for URI processing)

From: Simon St.Laurent <simonstl@simonstl.com>
Date: Fri, 26 May 2000 09:28:29 -0400
Message-Id: <200005261326.JAA30176@hesketh.net>
To: <xml-uri@w3.org>
At 05:43 AM 5/25/00 -0400, Tim Berners-Lee wrote:
>Oh good!  So Paul and Simon advocate that the XPath should
>be in the upper layer and therefore by Simon's model do
>absoltization.  Well, yes that would work fine. In fact, of course,
>the XPath spec does specify absolutization - and James Clark's
>implementation, I understand,  has a comment saying "fix this"
>where it ought to be in the code.
>
>The reason I'm on this list is that the xml plenary suggested
>that XPath should b changed to be in the bottom layer.

I'm worried that this discussion of layers isn't very nuanced, though it
may just be that you're discussing 'upper' and 'bottom' in terms of where
absolutization takes place.

I wrote a piece a year or so ago called 'Toward a Layered Model for XML':
http://www.simonstl.com/articles/layering/layered.htm

The gist of the piece was that XML 1.0 processing was weakly defined in the
spec and tended to glom a wide variety of different _kinds_ of processing
under the same header.  At the time, I was hoping that it might be possible
to bring namespace processing and DTD processing into some sort of
reasonable alignment, but that project neever seemed to take off in the
rush to schemas.

Similar issues have cropped up in further development, with questions like
schemas making entity-like contributions to the information set, where the
mechanism of XInclude belongs, how information that only has scope within a
portion of a document should be represented, etc.

The layers we've got are pretty thick, and it's clear from the inclusion
discussion that Namespaces in XML can't presently be treated as a simple
layer on top of XML 1.0 if cases like base URIs in external entities need
to be handled in what seems to be the proper manner (though there are cases
where no proper manner is clear).

Basically, we've got too much built too quickly without enough attention to
(or clean use of) integration techniques that seem pretty basic.  Issues
like the relative URI case seem intent on making layers that are already
quite large even larger.

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
Building XML Applications
Inside XML DTDs: Scientific and Technical
Cookies / Sharing Bandwidth
http://www.simonstl.com
Received on Friday, 26 May 2000 09:26:35 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:32:42 UTC