Re: [xsl] XSLT/XPath 2.0 (was "Identifying two tags...")

Subject: Re: [xsl] XSLT/XPath 2.0 (was "Identifying two tags...")
From: Jeni Tennison <jeni@xxxxxxxxxxxxxxxx>
Date: Tue, 14 May 2002 13:49:19 +0100
Hi Bryan,

> I think this touches on something I expressed a wish for some months
> back, although probably the expression was lacking in coherency what
> with my awful coffee problems, would it be possible to have the
> "processing-modes by which to control schema use (or more often
> useful, non-use)" be elastic enough that a vendor could specify what
> validation language they implement. I know this might lead to some
> fragmentation but I don't think it would be too terrible, obviously
> Microsoft would support Xml Schema, as would Saxon, Xalan and the
> major processors; however I think there might be a chance that Xalan
> would support Relax NG(seems like the Apache thing to do) and that
> would make me soooo happy(I've got a tear in my eye just thinking
> about this) I also expect that SUN would come with a processor that
> supported them all, and think about what weirdness might be possible
> if you had a specialized processor built on top of msxml coming from
> topologi or some such with support for schematron - oh man I feel
> like I've been doing helium!

When you say that you want support for RELAX NG and Schematron as
schema languages, what differences do you want validation against
those schema languages to make to XSLT processing? Do you simply want
your XSLT processor to validate against one of these prior to
processing? Do you want to get hold of the overall valid/invalid
result of that validation from within the stylesheet? Do you want that
record of the validation to be available at a low level (i.e. flags on
individual elements to say whether they're valid or not; Schematron
messages to indicate what the error is)? Do you want other information
that arises from validation, such as the name of the <def> that
defines the content of an element, or the type of the data in an
attribute, to be available?

Given that Schematron and RELAX NG purposefully don't supplement the
Infoset, I'm just wondering at the moment how useful it would be to
validate against them from within the stylesheet. I *feel* that they
should have equal standing with XML Schema, but given that they're
much more oriented towards *validation* rather than a kind of
*transformation* (which is what XML Schema does), I'm wavering...

> That said, is there a working Xml Schema for Xslt yet? James Clark
> suggested that one reason for Trex was that he couldn't find a way
> to represent Xslt in Xml Schema. If there is not one, doesn't it
> seem strange to base schema integration in a language with a type of
> schema that cannot describe that language, given that the language
> is an application of another language the schema format is meant to
> describe? What if I want to process Xslt using Xslt, do I derive any
> benefit? Not to say that I ever would with Relax NG but still.

XML Schema can't even validate itself :) That said, it's really quite
difficult to articulate every constraint on a particular markup
language, even with the best schema languages.



Jeni Tennison

 XSL-List info and archive:

Current Thread