Vista, etc.

[D. Hugh Redelmeier]

| From: Simon <simon80-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org>

| On 12/22/06, D. Hugh Redelmeier <hugh-pmF8o41NoarQT0dZR+AlfA at public.gmane.org> wrote:

| > I'm not even sure that X drivers don't compromise security: allowing
| > direct access to the video card may allow direct access to anything in
| > memory.  Even if you are only trusting the X server program, that is a
| > lot of code to trust.
| When you say X server program, what are you referring to?

For example, on my desktop (Fedora Core 5):
4     0  3322  3321  15   0 133308 90540 -      Ss+  tty7     551:24 X :0 -auth /home/hugh/.serverauth.3305

This is the thing that runs the display.

(An X client is typically an application program wanting stuff drawn on
the screen.)

| > | I
| > | also would say it got merged basically because it adds optional new
| > | functionality without breaking anything.
| >
| > Yikes.  As if X didn't already have too many features.
| 
| I don't think that better hardware acceleration support or input
| hotplugging  (an arbitrary alternate example) can be thought of as
| feature creep.

Almost everything is feature creep.  The question is: does this pay
for itself.  "it adds optional new functionality without breaking
anything" is an unconvincing justification.  There are probably better
ones, but I was reacting to what you said.

Building the X server on top of OpenGL is perhaps OK; certainly as an
option.  That is a fundamental change.  It seems to require much more
worthy OpenGL implementations.  It would seem to disadvantage old
desktops (without 3d hardware or enough RAM) and smaller systems (like
my Zaurus) and incent the use of closed source drivers.

The clients (like the flashy window managers) that use the 3d stuff
are what really forces the game.

Parallel example: X protocol used over the internet has gotten worse
and worse over the years.  The client programs and probably the
toolkits have not paid attention to parsimonious use of the network.
One of the worst trivial things seems to be enumeration of fonts.

FreeNX seems to be better over the wire, but that seems like a kludge.

| > Actually, my ongoing concern is that the state of open-source OpenGL
| > acceleration is really bad so requiring 3d performance practically
| > requires proprietary (closed-source) drivers.

| I agree about this, and it's enough to make me want to do something
| about it, because it's unconditionally removing all potential for
| Linux as platform for modern games.

What do you want to do?

Encourage vendors?  Worth doing, but progress seems pathetic.  I'd
like to better understand why.  ATI was open at one point.  So was
Matrox.  Not now.  Most convincing reason proposed: protecting
"Intellectual Property" licensed from third parties (or even: not
licensed).

Reverse engineer the hardware?  Very hard and laborious.  There is
real progress here on the ATI/AMD front (consider joining this
project).  I imagine that the worst part is trying to figure out how
to avoid hardware bugs.

Build open-source hardware?  I would expect that the performance of
GPUs built on FPGAs to be way behind what ATI and nVidia produce.
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists