On 12.03.2007 19:59:26 Abel Braaksma wrote: > Thomas Yip wrote: > > [...] > > <xsl:include href="../template1.xsl" /> > > <xsl:include href="template2.xsl"/> > > <xsl:include href="template3.xsl" /> > > <xsl:include href="template4.xsl" /> > > <xsl:include href="template5.xsl" /> > > > > [...] > > > > <!-- Body --> > > <fo:flow flow-name="xsl-region-body"> > > fo:block> > > <xsl:call-template name="template1" /> <!--in template1.xsl--> > > <xsl:call-template name="template2" /> <!--in template2.xsl--> > > <xsl:call-template name="template3" /> <!--in template3.xsl--> > > <xsl:call-template name="template4" /> <!--in template4.xsl--> > > <xsl:call-template name="template5" /> <!--in template5.xsl--> > > </fo:block> > > </fo:flow> > > [...] > > Now it is XSLT, not XSL-FO. I thought the XSL-FO process was causing the > trouble (see your own OP). > > I don't see why this (including many stylesheets) would cause you > problems with running XSLT.
I agree. Actually, the whole thing should be faster in a servlet since you don't have the VM warmup time. Plus in a servlet you can make sure you reuse the TransformerFactory and the FopFactory and finally reusing the JAXP Templates object representing the stylesheet could also help optimizing as the stylesheet wouldn't have to be parsed each time. Anyway, if the command-line is that much faster than inside the servlet the problem must be somewhere else: Either to little memory and the VM is constantly garbage collecting or some other thing (maybe like Andreas indicated) is wrong. I'd try to divide all the steps and measure them separately to find the culprit. Jeremias Maerki --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]