Before I spend a lot of time researching this, I just want to ask the community whether this would be possible: I like Linda tuplespaces a lot, especially when they're distributed and persistent. What I was thinking was that I could use ZEO (with a dbm storage) for this: the "out" operation would just write the tuple to the storage, while "in" and "rd" would try to grab a tuple, but if no matching tuples would be found the thread would be put on hold, and a special dummy Connection would listen to the invalidate messages back from ZEO (idea: subclass Connection, override invalidate()), trying to match the requests again on every promising ZEO update (for some value of promising). What to do with "eval" is something different, but I could envision some load-balancing scheduler here... Would this (ab)use of ZEO work? Any suggestions?