Re: protocol enhancements

> Sender: [email protected]
> From: David Dawes <[email protected]>
> Date: Mon, 24 Mar 2003 10:50:15 -0500
> To: [email protected]
> Subject: [forum] protocol enhancements
> -----
> On Mon, Mar 24, 2003 at 07:23:28AM -0500, Alan Cox wrote:
> >> committed to continuing to work on older X servers, and nobody is
> >> going to go the X12 route (break back compat). So new stuff is always
> >> either a library you can install regardless of your X server
> >> (e.g. fontconfig) or an extension that we use only optionally
> >> (e.g. RENDER).
> >
> >Both Gtk and X11 are going to break horribly in the next few years and
> >there is going to be a need for X.org, XFree or someone to actually fix
> >the standards to allow over 65535 pixels width and height. Otherwise you
> >are not going to be able to run X11 on your walls or on ultra fine
> >resolution displays.
> 
> Absolutely, and I'm glad to see someone else thinking about future
> hardware/technology.  This is an issue I've been trying to get some
> traction on internally, but it has been difficult.  For example,
> the responses I get (from Jim and Keith in particular) are like
> "that's not relevant, and won't be for 5-10 years", and "it will
> break the protocol", etc.  That's very short-sighted.  Xinerama
> with enough high-res screens can come close today (I can never
> remember if the limit is 32767 or 65535).  

In practice, the limit is 32768; the way we defined coodinates make accessing
bits above that hard.

For a major projector wall (which I've spent time looking into recently),
at 100DPI (the same sort of resolution we have today on most screens),
(about 8 meters, for the metric people).  The best walls I've seen to
date are at less than 1/2 this resolution (e.g. the Princeton wall),
and everything goes as the square of the resolution, so that is 4 times
as hard.

At a typical aspect ratio (say 3:1), at 100dpi resolution, the existing 
coordinate space will allow a wall of 26 feet to be built, and have 320 
million pixels, and require approaching 100 or more projectors, at a cost 
of >> ~100K (probably 1 million if you tried to build one of that scale 
today).  Building these is very much a research topic still, and X isn't 
yet a constraint.

So a guess of 5 years before it is a real problem affecting any significant 
number of people seems reasonable; 10 years is possible, but I don't think 
I said that.


> I'm sure that there'll
> be other issues exposed as we start working towards this.  I'd like
> to see at a minimum that the X server internals, and any proposed
> re-writes thereof in the context of 5.0, are prepared for this and
> for deeper and non-integer pixel types.  I also think that bumping
> the protocol minor revision should be explored as a backward-compatible
> mechanism for enhancing the core protocol.  For those who don't
> know, the X11 protocol has a so-far unused provision for a
> client/server protocol handshake, and this type of detail can be
> embedded in Xlib.  There's no need to go to "X12" to do this, or
> to keep tacking on miscellaneous extensions to deal with what are
> core protocol issues.
> 

Yes, the core protocol needs work.  Deeper pixels seem *far* more urgent 
to me than the coordinate space, and something most people will see much 
sooner than display walls of size and resolution that the coordinate space 
is an issue.  There, it is within the next doubling time: e.g. right around 
the corner. Most people already have 32 bits/pixel.

			- Jim


--
Jim Gettys
Cambridge Research Laboratory
HP Labs, Hewlett-Packard Company
[email protected]