RE: hardware xml / xslt

> Who is the fastest, Mike? Caucho's Resin? libxml/xsl? It is 
> not Saxon. Who do you think they should use?
> 
It's a long while since I've made any comparative measurements, and even
then, I never made them scientifically. I heard recently of one customer
site where putting in xt instead of Xalan gave a 12-fold improvement
(apparently Saxon broke, and I'm trying to find out why...). For the
last year or so, Saxon development has been focussed on conformance and
reliability (6.5.x) and XSLT 2.0 functionality (7.x), there has been
only marginal work on performance, and that has gone into optimizations
which probably deliver very large gains to a very small proportion of
users. 

I haven't measured any of the non-Java processors - it's really quite
difficult to set up a fair comparison.

I think that memory is probably a more critical measure than speed. For
a few sites, it's the number-of-small-transforms-per-second that's
critical, but most of the problems I get to see nowadays are large
document transformations, and they tend to be memory limited.

I'm personally rather sceptical of hardware solutions. Most of my
experience is that putting things in hardware can give a performance
boost at first (at the cost of very substantial investment in
development) but that software overtakes it in the end because the pace
of change is greater.

>I don't need to hear US-bashing on this list too.
> 
I made the factual comment that Datapower were a US software company,
with the unstated implication that you can therefore expect them to do
some aggressive marketing. I know many people who would take that as a
compliment.

Michael Kay


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list