There are a few new optimizations particularly in Saxon-SA, but for most workloads I wouldn't expect any dramatic difference.
This gives me the opportunity to announce to the list that I uploaded Saxon 8.9 yesterday, as those of you monitoring SourceForge will already know. I actually had to do a rebuild of the .NET version of the product because of one missing file, so if you've already downloaded it you may like to refresh it after a couple of days when the mirrors have settled down.
The big thing in this release is Java code generation - currently confined to XQuery only, and to the Java platform only. This is a Saxon-SA feature. Treat it as something of a beta release for the time being: there are a number of known restrictions, and the difficulty of achieving thorough test coverage means there's a higher-than-usual risk of bugs. I'm particularly interested in performance feedback: in most cases I'm seeing the elapsed time down to about 60% of the interpreted time, but it varies from 20% to 120%. (It should never be more than 100% and I will be giving those cases priority attention!). I hope to do something for XSLT and for the .NET platform in due course.
Apart from that, there are improvements in the join/index optimizations for Saxon-SA, particular in cases where no static type information is available.
Then as usual there's a long list of minor changes and I can only suggest you wade through the change log to see if your favourite wishes have been granted.
Michael Kay

From: [] On Behalf Of M. David Peterson
Sent: 13 February 2007 08:12
Subject: [saxon] Is Saxon on .NET 8.9 as fast as it seems?

In running some tests on the current site that I am developing, upgrading to the latest 8.9 release seems to have sped things up dramatically.   Just my imagination?


M. David Peterson |