If I had more energy, I might have considered trying to work NetBSD
   kernel concurrency.  But I suspect that is a major quagmire anyway,
   because fixing it requires such a huge code overhaul to introduce better
   kernel threading and eliminate the Giant Lock.


note that a lot of the kernel already runs OK without the big lock.
sure, we probably need things like interrupt threads and revamping
spl / lockmgr vs something else with real mutexes would be nice but
there's nothing really stopping you from working on some random
part of the kernel making *it* ok for properly threaded kernel...


i'm not trying to say it isn't a huge job to complete or that we
have all the primatives we want ... but that working on it is not
impossible.  :-)


.mrg.