Re: Module Proposal: GNOME Shell

[Owen Taylor <otaylor redhat com>]

Alan Cox wrote:

> I don't believe that is correct for any of the listed vendors even on
> Linux. On BSD the situation is even more patchy.
> 
> Is Gnome dropping support for these operating systems ?

The vast majority of GNOME desktop users are on Linux. The vast majority
of our developers are on Linux. In both cases we are talking more 95%.
I would supect more than 99%.

I don't consider saying "you have to have 3D graphics with modern memory
management" to be dropping support for BSD or other operating systems.
The hard part of the work - the design of the memory management and
drivers, figuring out how to program the GPU, etc, has all been done on
Linux and can be carried straight across, if anybody is sufficiently
motivated.

There is sadness in making it harder to get from bare metal to a fully
working desktop. But that's just the way computers are designed this way
- they have a CPU and they have a GPU, and if you want to create a
competitive desktop, you have to use both of them, not just one.

[ It's certainly *possible* to write code to render what we do in GNOME
  Shell completely in software on a fast multi-core machine, and get
  an OK framerate. GPUs of 5 years ago don't outclass modern CPUs that
  much. But it doesn't seem like a particularly interesting use of
  resources - either developer or CPU ]

> I'm not trying to start a fight or a flamewar, I'm genuinely interested
> in the directions as the board see it, especially given the hardware
> situation is not as was stated.

I can't really comment on how the board sees it.

> Even on Linux
>  Intel - no psb/mrst support (eg Dell mini 10)
>  Nvidia - many cards x86 only
>  AMD/Ati - similar situation, HD cards last time I checked not all
>  	supported

The statement that I made was:

 "Virtually all machines produced currently, or in the last 5 years
  have sufficiently powerful graphics to meet our needs."

Other people have made somewhat more sweeping claims; I certainly
wouldn't claim that every machine made in the last 5 years can be driven
appropriately with fully free drivers. It is a large fraction; and I
think we have good direction for much of the remainder.

Current problems we are having for people testing GNOME Shell are not
generally that drivers aren't there, but rather that they are buggy or
that we are triggering performance problems. Code paths have only been
tested with Compiz or haven't been tested at all. Giving the developers
of the graphics stack a simple straightforward target - this is how
GNOME 3 works - will help get these things resolved faster.

> Is this "the xyz is still supported in" seen as a short term thing where
> the panel rapidly gets deprecated/misses features or a long term path ?

As this thread shows, there clearly is not 100% consensus in this area
yet. But there is no development to update the panel to match the GNOME
Shell look or features. So, in my opinion, while the panel will be
maintained as long as people are interested in doing the work to
maintain it, it's not a GNOME 3 solution, it's a solution for people who
can't run GNOME 3.

- Owen