I have noticed 2 problems with the DumpHTML extension in combination with the latest version of mediawiki. Mediawiki version: 1.20.3. Version of DumpHTML: rev 115794 from mediawiki svn.
The first problem is that if you use 'subpages' ( /s in urls ) then the links are broken in the static output. The generated pages have the incorrect number of ".." relative links in them. This may potentially be resolved by setting 'munge-title' to md5; to prevent the extra slashes from being dumped into the name, but I didn't try to do so. ( and I don't want to do so, I want the titles to be left alone, not converted to a hash )
How come there is no way to output pages in the same structure as they exist originally? Why are you -forced- to use the new 3 folder deep "hashing" mechanism? I assume this is to prevent too many files from being dumpdr into the same folder, but there should be a way to shut this off and the code should be fixed to allow subpages.
The second problem is that image-snapshot seems unable to handle png images. The main icon in the upper left of my wiki is a png. A file for that is created in the duplicate, but it is a 0 byte file. The original file is not copied. It seems the images are not directly copied, and that DumpHTML is not able to handle pngs.