good point on AGPLed .emacs!  any AGPLed server running on Emacs to test it? ;-)

code vs config/data is indeed an issue, but I don't see that it has to do with P2P vs client/server, nor does it seem a particularly difficult one to me; it's just a matter of drawing the line.  many servers have config files in which you assign a database server password to a variable, but that shouldn't be regarded as part of the program.  conversely, there's plain data that *should* be regarded as part of the program (icons, for example)

what's different in the P2P case is that in order to interact with a peer you probably already have some implementation of the protocol.  should this imply you don't need access to other peer's improvements?  is it still your computing, when it's performed collectively by a set of peers?  or is it collective computing?  in which case, should every party be able to have at least some control over the partial computations performed by each party?  or should control be collective as in each party controls the part they perform?  can we even expect the answers to be the same for any P2P protocol/application?  I haven't given much thought to it, I'm afraid, and I'm glad you seem to have, even though your arguments seem to be more grounded on practical matters than ethical ones