Re: [xslt] Benchmarking

[Daniel Veillard <veillard redhat com>]

On Mon, Oct 22, 2001 at 03:35:49PM +0000, Bjorn Reese wrote:
> Robert Koberg wrote:
> > 
> > > Yes, we know. The XSLTMark test suite was used during development,
> > > and is part of the libxslt distribution (see test/XSLTMark). IIRC,
> > > our goal was to do a good job on benchmarks that really matters. We
> > > did not estimate that recursion was worth spending a lot of effort
> > > on.
> > 
> > Recursion is VERY important in XSLT.  Most experienced XSLT developers would
> > recurse rather than iterate. That being said, it is still pretty fast.
> > Recursion IS worth spending time on! Why do you say it is not?
> 
> I was speaking of the trade-off decisions made while developing the
> processor (note the past tense). At the time standard compliancy was
> more important than speed, pattern matching speed was more important
> than recursion speed, etc. We had a lot of ground to cover, and
> recursion was low-priority. Things may have changed since then, and

It may be surprizing but libxslt was not designed with performances in mind !

> I am certain that Daniel will appreciate any contributions that
> optimizes any part of libxslt, including recursion.

 yes. that said it's a very sensitive area, I appreciate patches, but they
would have to get a lot of scrutinity before going in :-)
 there is a lot of things I know could be improved, but as long as I get
real bug to fix, they have priority.

Daniel

-- 
Daniel Veillard      | Red Hat Network http://redhat.com/products/network/
veillard@redhat.com  | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/