Re: GDate API Rename Patch #2

[Sebastian Wilhelmi <wilhelmi ira uka de>]

Hi Havoc,

> I think better than runtime messages is to just turn off compat.h by
> default. The question is when do you do that.
> 
> The "deprecation policy" I wrote up a while back specified when;
> actually compat.h didn't fit in to that scheme, because it's more
> fine-grained than compat or no compat. But Owen didn't like the policy
> because it took like 3 releases to get rid of stuff (= 6-8 years at
> current release speeds...)
> 
> So I think we need a simpler, faster-to-deprecate policy that says
> when stuff gets deprecated, then implement accordingly.
> 
> gcompat.h seems sane though, the only questions are whether
> DISABLE_COMPAT_H is on or off by default, and maybe we want different
> variables than DISABLE_COMPAT_H (to disable/enable particular kinds of
> deprecated stuff).

If you ask me, I always prefer breaking compatibility if that helps the API.
Linux works the same way and I think it's better in the long run. Document the
changes, but remove deprecated functions as soon as possible, which means at a
major release, which seems to be about to happen for GLib soon. But of course
I'm not deciding. 

But at least I would turn compat.h off by default. On the other hand, looking
at the API changes GTK+ already contains I would very much like to just delete
the old cruft and require the (already big porting effort) to be a little bit
bigger. Not nice of me? Possibly. I won't raise the
everything-is-backwards-compatible-(in-theory)-thing of Windows to show
negative examples of overdone backward compatibility .....

Anyway, do as you please, I'm not feeling strongly about that, at least not as
strong as about the g_malloc virtualisation, which still makes me shudder ;-)

Bye,
Sebastian

P.S.: I'm off for X-mas holiday tomorrow. Wish you all a merry christmas and a
      happy release-rich ;-) new year 2001. 

P.P.S: Did I mention, that a new century is about to begin? Will it make it 
       into the history books as the GTK+-century? Who knows ;-)
-- 
Sebastian Wilhelmi
mailto:wilhelmi ira uka de
http://goethe.ira.uka.de/~wilhelmi