I just ran into this massive exerpt from the O'Reilly Java and XSLT book. Wow, if this is just one chapter, the book must be incredibly detailed... I'm going to put it on my list.
I'll have to see if it's on Safari as well to get a longer taste... I know Safari needs to make some artificial barriers so that you can't copy an entire book in 16 pages (one per chapter) but it's *so* nice to read the info like this article, presented as one long page - no download pauses between small sections (though this page is so big it actually slows Mozilla down - I've had to close the page in order to write this!). I'll have to try turning on background download in Moz and see if that helps.
Anyways, this chapter goes into details on how the different XSLT Java parsers work, how to use JAXP instead, and some details at the end on practical cacheing of XSL transforms (with a nice short singleton cache class that I've immediately copied for my own use).