
* do performance testing over the network
* do portable orphan timeout (we now depend on linux' select() behaviour)
* make network send/recv able to work on big endian
* default the number of threads to the number of cpus
* a library that is pthread-alike to ease replacing pthreads calls
* implement redundancy (should be only a couple of hooks in the networking
	code and new net ops)
* more documentation

