This contains several changes.
The XML format and naming conventions have been tweaked to meet
established Wayland conventions. I also updated the docs (they were
outdated in several places) and clarified more of the edge cases.
The surface_info event has been split into new_surface and new_output.
The former is now able to describe a hiearchy of surfaces and can be
used for subsurfaces, popups, etc. It is also used for hardware cursors.
The capturing client is now responsible for compositing these surfaces
if it desires, or omitting subsurfaces, popups, cursors, etc if it
prefers to. The new_output event is now tied to wl_output, and the
capture requests are combined into one.
The frame interface has been tweaked to support using a single frame
resource to receive several frames over the lifetime of your capture. It
seems like we were already moving in this direction, but some details
were lacking.
The frame interface has also been tweaked to more closely resemble
dmabuf import, but backwards. Despite how some people may feel about
the dmabuf import protocol, I think it's better to be consistent across
Wayland than to push that agenda. Regardless, most of the crappier
details of that protocol were left behind. The most important changes in
this respect were to change how planes were sent and described and to
unify the naming and language of the protocols.
Something I have left out is what to do about wl_subsurface. The
new_surface::parent argument should be limited to wl_surfaces like
xdg_popups. wl_subsurfaces behave differently enough that I want to
deal with them in a separate patch.
I also still have to write up the protocol error enums, but I'm lazy and
will do it after hearing some feedback on these changes.
