RE: What are the best XSLT 1 or XSLT 2 performance be

The benchmark contains about 30 (xslt,xml) pairs, together with reference
output for the transformation: some of them might have been challenging in
2001 but they are pretty small fry nowadays. The largest source document is
2Mb. Some of them are very untypical of user workloads. There are drivers
for a number of XSLT engines and an API for writing new drivers (in Java or
C++), and there is code to run a driver against all the tests (or a subset
controlled by a configuration file) and output the measurements. The driver
compiles the stylesheet once and then executes the transformation many
times, building the input document from raw XML each time (which means that
many of the transformations are dominated by parsing speed).

Regards,

Michael Kay
http://www.saxonica.com/
http://twitter.com/michaelhkay 

> 
>  From what I read about it, XSLTMark was just that: a set of 
> rather computationally intensive XSLT files that were to be 
> processed. Perhaps there was a little application around it 
> that measured the time or something. Some XSLTMark files are 
> also still present in the older sources of Transformiix 
> (Firefox's XSLT engine).
> 
> The Web Archive still has the original XSLTMark pages, but 
> unfortunately, they erased the binary files: 
> http://web.archive.org/web/20020611085801/www.datapower.com/XS
> LTMark/download/
> 
> Perhaps someone still has an original copy? If I understand 
> Michael correctly, unmodified, the files are allowed to be 
> distributed...
> 
> Abel
> 
> James A. Robinson wrote:
> > I've never used XSLTMark, but I see that libxslt 1.1.22 has in its 
> > source tree a path 'tests/XSLTMark/' filled with example inputs, 
> > stylesheets, and what look to be reference outputs.  Is 
> that what folks are after?
> >
> >   
> > 
> http://ftp.gnome.org/pub/gnome/sources/libxslt/1.1/libxslt-1.1.22.tar.
> > gz
> >
> > Jim