RE: [xsl] Anyone have benchmarks for msxml4?

Subject: RE: [xsl] Anyone have benchmarks for msxml4?
From: "Stuart Celarier" <stuart@xxxxxxxxxxx>
Date: Wed, 8 May 2002 11:38:46 -0700
Microsoft claims a typical 4x improvement of MSXML 4.0 over 3.0, and in
some scenarios an 8x improvement. Source: MSXML 4.0 release notes.

You have got to use caution in trying to apply benchmark statistics.
XSLT can be used for an incredibly wide variety of tasks, what do you
suppose should be measured? What's average or the norm? (Answer: this is
begging the question; there is no norm.) One person's performance
measurements may have so little to do with the your performance so that
at best they are meaningless, and at worst they are misleading.

It is more important for you to measure the performance on your
transformations, with your data, on your computers. The complexity of
the transform, the size of the data set, the amount and type of memory
and speed of the processor and hard disk (for paging) will have a lot of
influence your performance results.

Assuming that both MSXML 3.0 and 4.0 are installed, the task of
switching between the two amounts to changing the class from
Msxml2.DOMDocument.3.0 to Msxml2.DOMDocument.4.0 for the source and
stylesheet document objects.

Considering the benefits of compiling a stylesheet in an XSLTemplate
object should be done by thinking about how frequently the
transformation is used, and how many other objects are also being cached
in memory. It is a very dynamic situation. If you are trying to use an
XSLT solution in ASP or other middle-tier software, caching frequently
used or complex stylesheets will be an absolute necessity for
performance. Again, this situation is going to be different for
everyone: what's the work load on the servers, what are the processing
throughput requirements? You will find some discussion about the
benefits of XSLTemplates and sample code in my article at

If you have large source documents, you might well consider trying XSLT
processors that do not require the source document as a DOM. The DOM
representation of an XML document is commonly about 10 times the size of
the serialized XML document, and you can see that the DOM is impractical
for large amounts of data. The cost of parsing the XML into a DOM has
the potential to completely dominate the total cost of the
transformation. So look at XSLT processors that accept the source
document as a SAX2 stream, or the System.Xml.Xsl.XslTransform class in
.NET which can take an XPathNavigator or IXPathNavigable object in order
to avoid constructing a huge DOM in memory.


-----Original Message-----
From: owner-xsl-list@xxxxxxxxxxxxxxxxxxxxxx
[mailto:owner-xsl-list@xxxxxxxxxxxxxxxxxxxxxx] On Behalf Of bryan

I was just wondering if anyone had benchmarks on  how much processing
was improved using msxml4 over msxml3, and how great the improvement was
using xslTemplate and caching. Aside from that comparisons of xml/xsl
with ado would be cool, if anyone can point at a reasonable location.
Did searches on google, a lot of noise, most pointing at xslMark, which
I don't want to download and run just to get this info at the moment.

 XSL-List info and archive:

Current Thread