Re: Side issues [Was: Decisions we didn't intend to make]

[Federico Mena Quintero <federico ximian com>]

On Tue, 2006-08-01 at 20:35 -0700, Jeff Waugh wrote:

> > "Should we allow apps which use the language bindings in the Desktop" is
> > a valid question, though a very narrow-minded one if you happen to make
> > it at all ;)  Not allowing the use of bindings in the desktop is more or
> > less the same as saying that we don't dogfood our software.
> 
> That's a very fundamentalist approach to the question, and not one that I
> think has practical value. If we end up having ten languages/platforms in
> our Bindings release (ie. fully supported), should we accept software using
> all ten of those languages/platforms in the Desktop suite? That ends up
> being the important question here. Not whether we should have some generic
> inarguable policy, but the practicality of having the central components of
> our software maintained using ten different languages and platforms.
> 
> VERY IMPORTANT NOTE: This is an important discussion to have, but it is not
> related to the discussion we are currently having. Let's not get caught up
> on side issues.

Sorry, you lost me.  What are we discussing, then?

If we are discussing "gtk-sharp in bindings?  tomboy in desktop?" only,
then it's easy to consider all the combinations, since there's just four
of them.

I'm trying to make several points here:

1. In 5 years we'll have another language/VM/silo that all the cool kids
want to use.  We'll have exactly the same kind of long argument unless
we lay down some good rules *right now*.

2. I want the rules to be based on objective criteria so that the amount
of argumentation is reduced.  Also, I want the rules to be practical,
and to use the lessons from the past.

Guile-gnome was a well-maintained binding with zero applications written
for it, so when the maintainer disappeared, it just died.

Gnome-ObjC (Elliot's version) was pretty well-maintained and it had
three very small, mostly useless programs written with it.  When Elliot
disappeared it turned out that nobody was using those programs, so the
binding died along with them.

Are there any really good apps written with Java-Gnome, or Perl-Gnome,
or gtkmm?  This is an honest question:  maybe they exist but they are
not well-publicized.  [I don't know what Glom uses; if it is gtkmm, then
ignore the question for gtkmm.  [Note that I'm calling Glom an excellent
app here, and thus I really don't care what it's written in.]]

3. Maybe we should a) kill the bindings set, and say that apps that need
bindings are just using external dependencies; and b) declare that
programs in the desktop suite can *only* depend on API/ABI stable stuff,
even if the schedule for that stuff is not the same as ours.  Good
bindings for productive languages will bubble to the top, and if they
are sane they'll naturally want to align to our platform's schedule.

4. Having to maintain stuff in multiple languages.  This is not really
an issue, I think.  Most people don't hack on more than one or two
projects at a time.  People do need to learn various languages; that
will make them better programmers.  We only run into trouble if we make
the same mistake of the past, where we include very crappy apps written
with non-popular languages/bindings, such as boring/unpolished games
written in ObjC.

> Integration points as well as APIs. It's not something we've concentrated on
> much at all in the past, and is cause for some concern. I think this is a
> much more useful way to recast the point behind the Desktop suite, so we can
> actually have a healthy ecosystem of software around it instead of the inane
> ONE TRUE WAY we're stuck with at the moment (inclusion in the Desktop suite,
> because that's all there is).

You lost me again.  Can you be more specific --- what are the
integration points, or what would make a healthy ecosystem around the
desktop?

  Federico