- use software volume when hardware doesn't support all channels (alsa done)
- silence generation should be moved into the core to avoid races and code
duplication in the backends
-- rework resampler to not use pa_xrealloc. Use pa_memblock_new instead.
-- allow disabling shm in both client and server
+- pdispatch: split large memory blocks when sending
+- don't read/write audio data from/to ALSA devices if noone is listening
+
+Test:
+- autoload
+- sample upload + SHM
+- SHM revoke stuff
Auth/Crypto:
- ssl