I've tried this and the typical default download is an XML file in a very simple format (<file> and <dir> tags instead of HTML links). Obviously wget can't(?) parse XML files to get the links so it can't recurse.

So yesterday I wrote a simple Perl module which mirrors the directory tree into the current directory by parsing and extracting the links. Works nicely for the trunk I was trying to get but I'm not sure how robust it is. I'll try it out on a few more web trunks and if it's general enough I will submit it to CPAN