Monday, 17 June 2013

Re: non-Unity flavours and Mir

2013/6/18 Jono Bacon <[email protected]>:
> On Mon, Jun 17, 2013 at 3:39 PM, Scott Kitterman <[email protected]>
> wrote:
>>
>> The issue isn't that Canonical engineers aren't willing to work with other
>> people on integrating Mir, it's that because Mir is Ubuntu unique, has no
>> stable API/ABI, conflicts with other priorities, etc., integrating Mir is
>> simply not an interesting prospect for upstreams.
>>
>
> Mir is only going to be unique if upstreams and flavors refuse to consider
> it as an option.
>
> From what I am reading here Thomas has been very open to discussing with
> upstreams (both KDE and GNOME, and I am sure others are welcome too) how to
> build Mir support. We have also worked hard to be very open and transparent
> with Mir - regular status updates, discussion in #ubuntu-mir and the mailing
> list, and plentiful documentation for participating in the project. I
> consider that we are as open a project as Wayland and while you might use
> the "broad distro support" argument, Wayland hasn't shipped on a distro
> other than Rebecca Black Linux as far as I am aware. :-)
>
> I fully understand if you don't want to work on this problem, and I also
> fully understand if the KWin maintainer is uninterested in solving this
> problem and would prefer to focus on Wayland, but we are doing our best to
> be as open and collaborative as possible here, given the original points
> raised in Jonathan's email.
>
> I see this as a trade-off.
Fair point. But you can not expect KDE or GNOME to suddenly jump on
the Mir train. Supporting a new display server is pretty damn hard, it
took a lot of time to
clean up all the code to abstract X dependencies and make the switch
to a non-X displayserver possible. But after that is done, maintaining
a new display server backend is still not easy. And KDE and GNOME have
already put lots of work into Wayland porting, so why should everyone
now switch to Mir, a newly created project without stable API/ABI and
no obvious benefits except for Ubuntu support?
So, investing time into writing code for Mir makes no sense at all
from this point of view. There is still the option that Canonical
contributes code for Mir to e.g. KWin. Adding Mir support to KWin in
addition to Wayland would still be tricky, and doing it right be a lot
of work. Also, it would make code maintenance more complicated, so it
is understandable that upstream would reject the patches (similar
policy is applied for systemd to keep the codebase clean), as well as
I don't see Canonical writing patches for KWin/KDE - it just doesn't
make sense for the company to invest money into something which isn't
their primary product.
So I am afraid that this situation cannot be easily solved, and that
"you don't contribute to my project" accusations from any side will
not help solving it too.

> Jonathan raised a valid point about KDE's needs and made it clear that the
> Kubuntu team would prefer not to have to maintain Wayland as a foundational
> piece in order to deliver Kubuntu. Obviously this work can be performed by
> the Kubuntu team (or anyone else) if they wish to do so; the archive
> welcomes components that don't serve Canonical's needs.
>
> Canonical will of course be maintaining Mir as a core piece of
> infrastructure in the archive, and arguably encouraging Mir support in
> upstream KDE will help to alleviate this issue, but given that the Mir team
> are very open to supporting this outcome but both yourself and Jonathan are
> resistant to this, I am not sure what other options there are.
It is not about people being resistant. It is simple technical reasons
which make it difficult to resolve this situation.
In future, GNOME and KDE will share Wayland as display server, and
Ubuntu/Unity will have Mir, effectively breaking the infrastructure
unity we had between desktops apart.
From the current Mir specs, I also fear that replacing mir will be
easily possible (it is started at early boot and many things are
planned to connect to it as soon as possible - and Canonical would
have to think about a non-Mir situation when designing that
infrastructure - not sure if that will be done) But that's a different
topic.

Cheers,
Matthias

> What I am
> certain of is that Unity switching to Wayland is not an option, and
> Canonical is unlikely to invest in maintaining Wayland in the archive if it
> doesn't serve our needs in Ubuntu.

--
ubuntu-devel mailing list
[email protected]
Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel