Re:BTW, ODF is a file format (Score 0)

It is also likely that OOo is not a very good implementation of ODF with respect to reading and writing. It probably just unzips the whole file, parses it, puts it into its working data structures, and displays it; writing is probably the same thing in reverse.

I'm pretty sure that we will see something like super optimized zip-fs libraries for reading and writing that would allow fast update of just some bytes in the files.
I can also think of a zip-xml parser, where the xml in memory is still compressed. It would read a block of the zip file into memory and uncompressed/recompress on the fly. Modifying the ODF file would be as fast as modifying a binary file.