Lists Home |
Date Index |
Hey, thanks for all this Chris! I appreciate the time you spent to
help me/anyone else paying attention understand this better...
On Tue, 29 Mar 2005 12:42:16 +0100, Chris Burdess <firstname.lastname@example.org> wrote:
> M. David Peterson wrote:
> > Oh, I dont... sorry, didnt mean to lead you into the belief that this
> > is what I was suggesting. Your earlier numbers were significant and
> > while obviously the difference comes 100% from the transformation
> > engine at the time of the comment I simply wondered if you had found
> > ways to gain greater performance, possibly by illiminating, or better
> > said, integrating the libxmlj code more directly into the JAXP cose
> > base itself... e.g. Instead of simply providing a Factory instance
> > for JAXP to pass the proper information to binding the libxmlj library
> > directly into the JAXP code base, and fif so possibly adding a bit of
> > a performance enhancment.
> The figures given were for the 100% Java transformer, not the libxmlj
> one. The libxmlj transformer is considerably faster as long as it's
> working with libxmlj sources.
> The performance difference is down to the transformer rather than its
> integration into the JAXP framework. The same SAX parser and DOM
> implementation were used in both instances (GNU JAXP and Xalan).
> Both the 100% Java and the libxmlj implementations are integrated into
> the JAXP framework in a standard way and can be used with Sun's JAXP
> Chris Burdess
:: M. David Peterson ::
XML & XML Transformations, C#, .NET, and Functional Languages Specialist