> I haven't had a chance to analyze this yet (and I may not be able to
> this week), but one quick point is that wherever the Saxon performance
> problem is, it's not actually in exsl:node-set(). That's because all
> that exsl:node-set() does is to flip a bit saying that the tree can be
> used as a normal document.
> 
> The problem (or opportunity for improvement!) is much more likely to be
> in optimization of copying operations on trees.
> 
> An observation, on line 65 of incelim.xsl
> 
> 	<xsl:copy-of select="exsl:node-set($rng)"/>
> 
> there is no need for a call on exsl:node-set() here. You can copy an RTF
> directly.

Michael,

I've followed your suggestion and fixed the typo -- it was an oversight on my 
side instead. Note, though, that this operation is only executed once
per the stylesheet, and as you said, the only thing node-set() does is to
flip a bit.

The execution time didn't change at all, and it should not change.

David

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