Screen shooting and recording protocols (Re: Authorized clients)

[Bill Spitzak]

I'm pretty certain what is being requested is a rectangle of some 
output. The video recorder will record whatever pixels are inside that 
rectangle, even if the surface used to choose it is moved away from it 
or if other surfaces enter it.

The compositor must be able to set a stride value on the buffer so it 
can send the cropped rectangle from a buffer that is the size of the output.

I think "recording a window" sounds a lot like another output (as in a 
montior). Then all the existing multi-view stuff can be reused to 
duplicate the window into this output, including controlling whether 
child surfaces are shown.

This will allow showing of subwindows, which I think may be impossible 
for a client to do if it launched a child that is drawing the subwindow (?).

Maybe when a video recording client starts it can advertise this 
additional output. Then the client has to do something to clone it's 
surfaces onto this output. And the surfaces will keep updating even if 
the user hides the original window.