RE: XSLT with SAX, Reduced Instruction Set Optimization and Liter al Result Element Stylesheets.

[Jose Alberto Fernandez <JF...@viquity.com>]

> -----Original Message-----
> From: Joseph_Kesselman@lotus.com [mailto:Joseph_Kesselman@lotus.com]
> 
> 
> >What strucks me as quite strange is that the XSLT spec does 
> not provide
> for
> >specifying an input schema
> 
> Not all documents have schemas; XSLT wants to be able to process
> well-formed documents as well as fully-validated ones.
> 

I have no problem with XSLT's goal, but that does not mean that XSLT shall
not use schema info if available. This is even more striking given that fact
that we have I do not know how many groups defining very detailed DTDs for
all kinds of things and XSLT is not able to use any of that.

I which the spec had defined some optional entity:

	<xsl:input-dtd .... />

so that processors could decide to use it for optimizatons.

> >or for providing a way to secify "select THE ONE
> >AND ONLY ONE firstname in the person entity"
> 
> I think that can be expressed with "firstname[1]". Part of 
> the answer to
> streaming may be to ask that stylesheet authors make their 
> intentions more
> explicit.
> 
> In fact, specifying [1] when you know only one is expected might yield
> performance improvements even in non-steaming 
> implementations, if they're
> clever enough to realize that this predicate will succeed only once.
> 

This is a very good tip. Do we know how Xalan would behave with this?

Actually, if by looking to the schema the compiler where intelligent
enough to introduce this kinds of markings, this could be an easy way
to optimize a stylesheet.

Comments?
> 
>