Re: Time for an exslt for 2.0?

>>>>> "David" == M David Peterson <m.david.x2x2x@xxxxxxxxx> writes:

    David> Are you thinking along the same lines in which the original
    David> EXSLT project (assuming that my understanding as to why the
    David> original project was started) was focused towards, in
    David> essence adding to the mix the instructions and functions
    David> that were left out of the spec for various reasons or had
    David> since been realized as necessary?

Not necessarily instructions and functions - anything that promotes
writing portable stylesheets.
The two issue I mentioned are cases in point.
Dmitre added saxon:memo-function="yes" attributes to some of his
prime-number-calculating functions. When I looked at these, I promptly
realised the benefit, and implemented my own attribute in gexslt with
identical semantics. But this meant coding BOTH attributes within the
function definition. As Dmitre pointed out, for two processors this is
just about OK, but if more and more implementations were to do the
same thing, it would be frightfully messy to read, and a real pain to
have to keep adding new attributes for each new processor.
And having a standard way of accessing environment variables is
another pure gain on portability (otherwise, if you need this
facility, yopu are going to have to do a lot of unecessary coding with
xsl:use-when - it's possible, but a nightmare for maintenance).

Then there are things in the XSLT 2.0 spec. that are left entirely to
the implementation (such as collation names - this is actually under
discussion on the qt-comments list at the moment). It might be useful
to have a set of standard collation names with known properties.
-- 
Colin Adams
Preston Lancashire