[RFC] wl_surface video protocol extension

[Daniel Stone]

Hi,

On 16 October 2013 19:40, Jonas Kulla <nyocurio at gmail.com> wrote:
> 2013/10/16 Daniel Stone <daniel at fooishbar.org>
>> On 16 October 2013 17:09, Jonas Kulla <nyocurio at gmail.com> wrote:
>> > it might just be a stupid thought of mine, but it would be kinda cool
>> > if there was wayland protocol for creating an EGLStream [1] backed
>> > surface.
>>
>> It would be nice, but TTBOMK the only EGLStream implementation is in
>> the Tegra EGL stack, which doesn't support Wayland ...
>
> Yeah, but would it make sense to write a completely new specification
> if there's already an existing and ARB approved one? What would prevent
> eg. Mesa from implementing EGLStream?

Sure, but by the same token we could've skipped wl_surface entirely
and used EGLStream there.  It's a pretty complex and untested
protocol, and given the wide lack of either adoption or interest (it's
been around quite some years now), I'm not really sure what the
benefit would be to not only leaping over there, but also requiring
all multimedia content to be slammed through EGL.

There's also the wider issue of EGL not having events (so not being
able to give feedback beyond your initial - likely wrong - estimate of
latency), of not being able to queue frames in EGLStreams but having
it be very immediately driven (so making it ever more likely that
you'll miss your frame deadline, and eliminating a whole class of
power savings), and more.

Cheers,
Daniel