[22:59] <oiaohm> http://dvdhrm.wordpress.com/2013/09/01/splitting-drm-and-kms-device-nodes/  I really do wonder badly this is going to effect mir idea of server side control.
[23:25] <RAOF> oiaohm: Not at all.
[23:26] <oiaohm> RAOF using device nodes could nicely under mine buffer controls.  Client side allocation in GPU memory spaces.
[23:27] <RAOF> They can't allocate the front or back buffers, which is all Mir controls anyway.
[23:29] <RAOF> We weren't planning to, and don't, allocate *all* client buffers. That would entail much annoyance - we'd need to have server-side changes any time a GPU driver wanted a differently laid out buffer for internal purposes.
[23:29] <RAOF> Clients allocate all their textures themselves.
[23:32] <oiaohm> RAOF: the back buffer one is something I do wonder about.  dma-buf  that is used to pass from device nodes back to servers can have a back buffer applied.
[23:32] <oiaohm> basically the change might effect you a little.  RAOF
[23:33] <RAOF> You're using "back buffer" in a way I don't recognise.
[23:33] <RAOF> dma-bufs are a buffer passing mechanism, yes.
[23:34] <RAOF> There's no way for a client to *display* anything except through Mir; only what they render to the buffers sent out from the server (as dma-bufs) will ever make it to the screen.
[23:34] <RAOF> (Barring hilarious client bugs)