[ubuntu-x] plans for Lucid?

[Bryce Harrington]

On Wed, Nov 18, 2009 at 08:13:35AM +0200, Timo Aaltonen wrote:
>    There haven't been any published plans for the X stack in Lucid, so 
> here's what I think of it.
> What do you think? What did I miss?

> xserver:
> 
>   Debian Squeeze will freeze in March, and having xserver 1.8 there is 
> possible, if someone will work on it. Currently they are understaffed (who 
> isn't :), which is why 1.7 has not seen much work lately. Having the same 
> stack in both Squeeze and Lucid should help in getting more widespread 
> testing, and share the devel effort.

As I see it, there are 3 choices with the following trade-offs:

1.6.x:  Known quantity, well tested, looks like it will continue being
supported for a while going forward.  Most video drivers will continue
supporting this version.  Cons: Lacks newer input device work, older
release.

1.7.x:  Released with one point release, maybe more.  Most video and
input drivers will support this version.

1.8:  Not released yet, but scheduled for release in December.  

There are a few factors which play into the decision:

1.  Lucid is LTS, so by default we would prefer a known-stable release
already in hand, unless a newer version looks like it will be supported
in debian or upstream for a longer period.

2.  What goes into Debian Testing.  (1.7 so far presumably.)

3.  For boot speed improvement it is desired to drop HAL.  We need HAL
for tablet support, so to avoid regression we need a udev-enabled wacom
driver.  xf86-input-wacom 1.x with udev will be supported only on 1.7
and newer, thus ruling out 1.6.

4.  We plan to place a focus on proprietary driver support improvement
this release.  Thus it would be beneficial to have -nvidia and -fglrx
functional as much as possible during the release, rather than block
waiting on rebuilds by the vendors.  This would suggest preferring 1.7
or earlier.

A couple caveats to #4: The drivers may end up blocking anyway due to
kernel version change.  An approach which has us on 1.7 (not a git
snapshot) up until when 1.8 is available, would provide a period where
the proprietary drivers work, for us to do packaging/testing.

Given all the above I'm leaning more towards xserver 1.7.  I know a lot
of new features are planned for 1.8, but don't know yet that those
features are absolutely required in Ubuntu.

> mesa:
> 
> 7.7 should be out by the end of year, so that should be an easy choice. 
> Might also make r6/7xx DRI work properly.

Agreed.  If 7.7 is out in a couple months it should be no problem to
include.  I would like to avoid having to pull in a mesa update at or
after beta, as we did for Karmic due to the risk of regression it might
incur.

> libs:
> 
> whatever are needed for xserver 1.8.
> 
> drivers:
> 
> - intel 2010Q1 release (might need mesa 7.8 and perhaps kernel 2.6.33?) - 

Sounds like the kernel will be 2.6.32.  Perhaps we should look to what
is released for 2009Q4?  jbarnes - advice?

> - xf86-input-wacom 1.x (new clean release with support for properties etc)

As mentioned above, for this we'll want 1.7 or newer to get the udev
support here so hal can be dropped.

> - rest according to what the xserver might need (evdev..)
> 
> kernel:
> 
> generally, we want the latest and the greatest.. AIUI it's not yet decided 
> if it's going to be 2.6.32 or .33.

I think latest and greatest is good for normal releases, but since this
is LTS I would really prefer something known to be stable and safe.

Bryce