[Libdlo] [RFC 2.6.30 1/1] defio fbdev displaylink implementation
bernie at berniethompson.com
Thu Oct 8 09:30:22 PDT 2009
Thanks for jumping in (and I hope we do get to implementation of some
of your earlier clear thinking about the right approaches for user
mode clients! )
Definitely agree about not getting distracted by the non-bottleneck
cases. One thing high on the list, though, is multiseat - so
individually you have a bunch of displays that wouldn't hit a
bottleneck (USB), but collectively they do.
But #2 has problems that way, anyway, as CPU & memory controller are
certainly also a bottleneck in the multiseat case, if you assume
people will use low-end servers (which they will, because of the focus
on cost). So obviously getting the best possible damage information
in a standard-as-possible way (e.g. #1) is the biggest possible win.
More information about the Libdlo