> > I have tested the XSL sheets with Instant Saxon and XSLTproc under
> > windows/cygwin. Both these took 25-28 minutes to complete to generate
> > the HTML Help output target of the XSL sheets.
>
> under linux (with ibm jdk-1.3.1) i had 8 minutes using saxon
> and 13 minutes usind xsltproc on a 500MHz pII
Instant Saxon uses MS JVM, XSLTProc is exe... My testmachine was
a 850Mhz AMD Duron.
> > The output size of XSLTproc was slightly bigger, as it added "
> > as &qout; in the HTML, and the HTML header part also differed.
>
> > A more critical thing is interlinked function references. In the
> > Saxon output, the functions were not linked, in the xsltproc
> > output, they were... Hm...
>
> haven't noticed that, will re-check
There can be other related problems...
> > I have also checked the class and function pages. Class dir
> > parameter list looked this way for both versions:
> >
> > dir {
> > dir(string directory);
> > string path ;
> > string read();
> > void rewind();
> > void close();
> > }
> >
> > Uhm, this is not what it should look like
>
> this is experimental stuff usind <classsynopsis>,
> still needs stylesheet customizations in both dsssl and xsl
Yes, I have just seen in the source, that the dir class was
changed... So this is not an error. I compared it to my latest
DSSSL builds, and as they were built before the change, this
file contained only one line for the constructor here...
> > I also checked array_keys, and the param list looked like this:
> >
> > libxslt: array array_keysarray input [, (mixed search_value )
> > saxon: array array_keys(array input [, mixed search_value )
> >
> > Interesting, but this is definetly not what I expected to see...
>
> no closing ] in both of them? -> have to re-check ...
In LibXSLT, the ( is also missing, and there are no spaces
between the funcname and first parameter. I guess this is
because there were spaces in the funcsynopsis version, and there
are no spaces in the source in the methodsyn version...
> > BTW. libxslt generated files in the directory where I started
> > it, with no respect to the XSL defined output dir, ./html
>
> yes, known problem ... Sebastian Bergmann reported that index.html
> is generated in the specified dir, but not the chunks
> will report that to the libxslt list today (hope so)
Actually the index file was named manual.html and not index.html
for both processor, which is an error two.
Goba
