Current thinking is to build a tiny little ZODB implementation in C (anybody have one?) to access the DB on the handheld, plugged into a tiny little HTTP server like thttpd or Abyss, and an embedded browser like ViewML to access it.
ZODB is pretty intimately connected to Python. Rewriting all of ZODB in C would be pointless, since you'd still need a Python interpreter to make sense of the objects coming out of the database.
Maybe you just need a simple data store, like a Berkeley DB file.
Sort of, but Berkeley DB is probably too big and complicated to use on a wimpy little handheld - the library alone is nearly 1Mb. Hmm. How about some sort of custom simple data store with a Python wrapper on it, so that I could use Zope to let people help build the content sets from wherever they are, and then just move the whole content wad to a handheld? The content's going to be something like simple HTML - like SquishDot articles - with a catalog holding metadata: language, subject, expiry date, source... Sort of like: [ Zope ] -> [ Python wrapper ] -> [ C API ] -> [ Content DB ] in the content management context and: [ Embedded HTTP server ] -> [ C API ] -> [ Content DB ] for accessing it. Hmm... Has anybody done anything like this?