Gnome-2.6, GStreamer-0.8 and caps [was: Re: [gst-devel] more caps]

[Thomas Vander Stichele]

El jue, 25-12-2003 a las 22:39, Colin Walters escribió:
> On Thu, 2003-12-25 at 09:56, Benjamin Otte wrote:
> 
> > I always work on the branch that is the future and not a dead end. I'm not
> > even really interested in backporting fixes. So if you do this, you'll
> > probably lose me as a developer on the Gnome 2.6 stuff.
> > Branching is never a good solution - it's just like forking, without that
> > bad name. 
> 
> Nah - it's a branch if stuff is being actively merged both ways.  It's a
> fork if not.

Branches work very well if the people doing the branch make it work
well.  It's an instrument, not an end goal.

> > We still have 2 months time to make all this work. 2 months ago we did
> > have neither navigation nor interfaces nor tags.
> 
> No...GStreamer doesn't exist in isolation.  If stuff only starts working
> in 2 months, then all the app writers who also want to get their
> applications in GNOME are going to be screwed, since they won't have a
> stable base to build on.

It won't come as a surprise to anyone that I feel the same way.  What
little apps out there that actually are foolish enough to use GStreamer,
aren't being treated well by us.  We're going at it as if the final
GNOME 2.6 release date is also our 0.8.0 release date.  Basically, all
of the apps that use GStreamer are now broken because either they depend
on the stable branch which everyone wanted to drop asap, or on HEAD
which means they have no more than 2 months left to get stable
themselves, but probably more like a week since it'll take us those 2
months to settle down.


> And there's still tons of other work we have to do like getting length
> metadata, and fixing all the bugs that are sure to come up with people's
> corrupted and crappy MP3s.
> 
> And that's just for Rhythmbox (my primary area of concern).
> 
> At the moment, I am willing to give Dave the benefit of the doubt.  But
> if we can't get things working again fairly well in the next 5 days or
> so (and I am willing to give it a damn good try!), we should think
> seriously about reverting back, as much as that would suck.
> 
> > After every huge merge there's huge breakage. Noone can make a branch
> > work perfectly on his own. 
> 
> Perfectly?  No...but verifying that major applications work would be a
> good indicator of whether a branch is mergeable.

I agree. Company, I agree with you that a lot has changed and a lot of
new exciting stuff has been added.  Only, the way it gets added seems to
be a bit uncontrolled.  While we might be laying good foundations for
the future, we're actually worse off atm.  There has got to be a middle
line between "I'm not changing all of this on a branch because then it
doesn't get done" and "Let's break HEAD for a few months since then
people will try to fix stuff".

Leaving what little apps we have out there (RB, n-m, s-j) completely
broken is not going to give people a better impression of us.  I don't
know if people notice, but apps are dropping GStreamer and switching to
xine all over.  And that's for simple audio playback !

I was sceptic of us being able to deliver a good 0.8 for GNOME 2.6  I'd
have never proposed it myself.  I was happy to see it get proposed
though by our core hackers, because that gave me the impression that
they were willing to spend time fixing direct end user problems.  I hope
that is still going to happen; if not we should pull out 0.8 from GNOME.

GStreamer has been around for 3 years or more now.  I think it's about
time we start getting some basics right as well.  Both in the code we
deliver as well in the way we deliver it.  If stuff can't be done on a
branch while people can fix their apps against HEAD in preparation for
GNOME 2.6, there is something wrong with what we're doing on the branch.

I think that given the fact that we don't have that many core hackers
around, we've bitten off a bit more than we can chew.  I see some value
in Ronald's idea; branching off from before caps for 0.8, but it all
depends on how quickly current caps can be stabilized.

In essence, our random-walking is getting us only slowly in the right
direction.  I think it is vital for GStreamer that it makes a solid
foundation for GNOME to build on.  If GStreamer can't do that, it misses
its entire point of being.  With some of the Helix discussions in GNOME
that went on, there's even a chance GStreamer will be dropped.  As a
GNOME developer, I can't even say that would be the wrong decision based
on what GStreamer has delivered to GNOME up to now.

I'm pretty sure that, OTOH, if GStreamer does get dropped from GNOME,
it'll pretty much be the end of it too.  Sure, the few of us can still
take it apart again and again and do our little things with it.  But it
won't grow into the big widely used framework we intend it to be, and it
will have missed its chance to make a mark.

It is not too late for us to fix things and get it right, but it will
require cooperation from everyone with a clear set of well-defined goals
and a decent game plan for 0.8/G2.6

BTW, I'm not blaming anyone; I know I've dropped the ball on things just
as well. This is a hard project to manage and as we said before we're
missing clear leadership.  But it's something we have to deal with.  I
don't think we can just keep pretending we know what we're doing,
rewrite stuff over and over, and hope it's going to turn out fine by the
date where it's due.

Thomas

Dave/Dina : future TV today ! - http://davedina.apestaart.org/
<-*- thomas (dot) apestaart (dot) org -*->
I'm so boring my clothes wanna keep
somebody else warm someone cooler
<-*- thomas (at) apestaart (dot) org -*->
URGent, best radio on the net - 24/7 ! - http://urgent.fm/