Saxon 6.x implements collation for xsl:sort using the Java collation rules for your locale, or for the selected language if you specify the lang attribute of xsl:sort. Saxon 8.x offers much more detailed control over collations, which is described at
http://www.saxonica.com/documentatio...collation.html
I'm afraid it's rather daunting technically - there's a lot of background to become familiar with, and the end result is that it can still be quite difficult to get exactly the collation you want, especially when it comes to handling of spaces and punctuation. But as far as accented characters are concerned, the collations for most European languages will sort first by the base letters, then by case, and then by accents, which means for example that Kuchen will sort after Küche - the umlaut is taken into account only if everything else in the word is identical. There are exceptions for some languages where certain "accented" letters are considered to be first-class letters in their own right, for example this applies to ø in Scandinavian languages.
The simplest way to find out what the software does is to experiment.
Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference