Where should project Weston go?

[Pekka Paalanen]

On Wed, 10 Dec 2014 12:36:19 -0800
Bryce Harrington <bryce at osg.samsung.com> wrote:

> On Wed, Dec 10, 2014 at 04:00:43PM +0000, Daniel Stone wrote:
> > A platform that people can develop
> > desktop environments on? Sure! I guess a lot of those would be fine as
> > Weston plugins, though ultimately we'd need something like libweston to
> > make that workable. My biggest problem with libweston as it stands is that
> > I think it stands to repeat the X.Org 'API' disaster: just copying all your
> > header files into a directory somewhere and installing a library isn't an
> > API. I'd like to have a much more serious look at what we expose, how, and
> > why.
> > 
> > That would be quite a bit of work, but I think far more valuable if it
> > allows us to get rid of large swathes of desktop-shell.
> 
> I think this is a really good point.  If we're going to be supporting
> libweston's API as a stable interface, then it deserves going over the
> interface (and overall scope) with a fine toothed comb before other
> projects start using it.
> 
> Also, changing the library name will be harder once the effort becomes
> public, so would be worth considering if "libweston" is the best name.

Yes, but I believe this is too huge a task to be done in one release
cycle. (I'll keep calling it libweston for the time being.)

Therefore I have been suggesting, that until we are really clear on the
APIs, we keep the current policy for ABI stability: every minor release
is allowed to break ABI. Only incremental micro-releases maintain
stability.

To make that work, I suggested that different minor versions of
libweston need to be perfectly parallel-installable, so that a
user/distro can install external projects that require different
versions of libweston at the same time without any pain.

Wouldn't this be sensible for at least middle-term? We would not need
to be hasty with the stabilization then.

If we ever can stabilize the libweston ABI, it might be worth calling
it Weston 2.0, perhaps? We could use that to signal that minor bumps do
not break the ABI anymore.


Thanks,
pq