Re: Gtk+4.0

[Chris Vine <chris cvine freeserve co uk>]

On Sun, 14 Aug 2016 22:48:24 +0100
Emmanuele Bassi <ebassi gmail com> wrote:
> On Sunday, 14 August 2016, Chris Vine <chris cvine freeserve co uk>
> wrote:
>
> > On Sun, 14 Aug 2016 21:22:06 +0200
> > Sébastien Wilmet <swilmet gnome org <javascript:;>> wrote:  
> > > On Sun, Aug 14, 2016 at 07:17:34PM +0100, Chris Vine wrote:  
> > > > On Sun, 14 Aug 2016 13:40:55 +0200
> > > > Sébastien Wilmet <swilmet gnome org <javascript:;>> wrote:  
> > > > > When GTK+ breaks the API, it doesn't mean that a higher-level
> > > > > library needs to break API too. For example, GtkTextView has a
> > > > > quite stable API, so I think GtkSourceView will still have a
> > > > > stable API too, to keep backward compatibility during
> > > > > GtkSourceView 4.  
> > > >
> > > > However, if your application depends on two GTK-based
> > > > libraries, the developers for one of which proceed to adopt a
> > > > development path using unstable GTK versions and one where they
> > > > stick to stable, surely you are doomed?  Likewise if the two
> > > > libraries decide to adopt different stable versions (not
> > > > impossible with a two-year cycle for stable releases).  
> > >
> > > Yes, that's why it's more important for libraries to follow
> > > unstable GTK. For apps, it's less important.
> > >
> > > GtkSourceView will probably follow unstable GTK, while still
> > > trying to keep the GtkSourceView 4 API stable.  
> >
> > I was not referring specifically to gnome based libraries.  There
> > are plenty of others, which will definitely not want to follow the
> > unstable series.  Even a two year cycle of stable GTK versions will
> > probably be problematic for them.
> >
> > It really comes down to the question of what GTK is.  If it is the
> > GnomeToolKit (or GTK developers are happy for it to become the the
> > GnomeToolKit) then I can see the merit in the proposal.  Otherwise
> > it looks to me like a suicide note.  
>
>
> So, let's not beat around the bush: GNOME developers are the vast
> majority of the GTK contributors, so clearly they get to steer the
> project the way they see fit. You get to steer the project only if
> you show up and do the work.

I agree: "... If it is the GnomeToolKit (or GTK developers are happy
for it to become the  GnomeToolKit) then I can see the merit in the
proposal".  It is, and should remain, a rule of open source development
by volunteers that those who write the code have the right to make the
strategic decsions.  If not, they can vote with their feet (or with
their fork).  Having said that, I suspect but do not know that a
proportion of those writing GTK code are employed to do so, so in their
cases I suppose it is the employer who calls the shots.  Obviously,
those who make the decisions must also accept the consequences of the
decisions they make.

> I'm more interested as to why you think this new release policy of an
> API and feature stable release every two years as a "suicide note"
> considering that the whole thing has been drawn to cater to non-GNOME
> consumers of the API after listening to their complaints. GNOME app
> developers are pretty much used to keep up with bi-yearly releases,
> whereas non GNOME app developers are often complaining about cycles
> that are too fast.
>
> If two years are still too fast we can definitely look into making
> API and feature stable cycles longer; that usually comes at a price
> of making porting more difficult, but if that is an acceptable cost
> we can definitely do feature-frozen releases every four years instead.

I fear I am not good at expressing myself.  I see the problems as: (1)
every GTK-using library that an application links to must link against
the same ABI version of GTK.  This means that all such libraries that
the application happens to use (as well as the application itself) must
go in step.  I accept that this should not be a problem if GTK becomes
the "gnome tool kit", as the gnome project can enforce its own internal
rules on this.  (2)  Bearing in mind this linking issue, I suspect a
2-year stable release cycle is likely to be too high a churn rate for
non-gnome applications and libraries.  The proposal is in my view
likely to cause non-gnome application and library developers to look
elsewhere.

I may be wrong on that.  You may be right.  Time will tell, if that is
the way it goes: this is a competitive market.

Aside from that, declaring GTK to be the "gnome tool kit" may serve to
add welcome clarity on the point.