I'm sure many of you have seen the slashdot story with the same title as this thread. It pointed to http://www.softlab.ntua.gr/~ttsiod/buildWikipediaOffline.html which is the description of a simple offline Wikipedia reader which runs off the bzipped dumps.
I for one thought the use of bzip2recover to seek over the bzipped dumps without decompressing the whole thing was genius. And I was even more excited when I found out how simple the source code to bzip2recover is so small (just one .c file with no includes or anything, see http://swtch.com/usr/local/plan9/src/cmd/bzip2/bzip2recover.c).
So it should be easy to tweak the bzip2recover program to eliminate the need to split the file up at all. I'm kind of surprised if there isn't already something out there to build and index to seek through bzip2 files, but I've been looking for one for quite some time and didn't find anything. Now I know it's possible though, and apparently not very difficult.