[Xcb] icccm

[Jamey Sharp]

On Mon, Mar 21, 2005 at 01:18:21PM +0100, Vincent Torri wrote:
> >Sure, you can queue the reply cookies that XCB gives you. If you can be
> >certain that you enqueue them in the order that XCB produces them, then
> >they'll be in the same order as the returning replies, so you can
> >minimize latency there. That's an easy invariant to maintain as long as
> >you have one queue per thread (which is obviously true if you're
> >single-threaded).
> 
> it would be interesting to do that for evas. But, i have to understand the 
> mechanism. Is it possible that someone writes a very simple program with 
> xlib, and another in xcb using the ssytem of queues, written such that the 
> differences are explicit ?

I don't think I can write it, because I'm not sure what you can do with
this queue mechanism that you can't do some more direct way. (I can
believe it's good for something: I just don't quite understand what.)

If somebody could point me at the architecture discussions that led to
this idea, and especially if somebody wrote a simple illustrative Xlib
app that they believe would benefit, I could try porting it to XCB and
illustrating use of a reply cookie queue.

--Jamey