Re: Say goodbye to core X fonts

[Owen Taylor <otaylor redhat com>]

On Wed, 2003-04-16 at 11:48, Arnaud Charlet wrote:
> > If you do some performance analysis of what the PangoX backend 
> > is doing and how much memory it takes to do it, you might feel a lot
> > better about that. :-)
> 
> I really hope so. Having real figures would certainly be helpful.
> Also, what is the disk size of a expat/FreeType/fontconfig/libXrender/Xft
> installation ?

Approximately:

 128492  /usr/lib/libexpat.so
 327490  /usr/lib/libfreetype.so
 148183  /usr/lib/libfontconfig.so.1.0
  25936  /usr/X11R6/lib/libXrender.so.1.2
  67950  /usr/X11R6/lib/libXft.so.2.1

So, a bit over half a meg.

> > Perhaps the right model here is the Java model of a 'SDK' and 
> > 'Runtime Environment', which people seem to deal with pretty
> > well.
> 
> Well actually having to install the right JVM is also a pain in the neck,
> so we usually avoid any Java based application :-)

I think we can avoid the multiple-slightly-incompatible-in-weird ways
problem of the JVM :-)

> But this general goal is certainly a nice one. I am just getting the
> impression that all these nice things tend to be simply ignored because of
> other issues, and therefore tend to never quite happen

I think rather it's simply that the issues in this area are:
 
 A) Not of personal (or professional) interest to the majority
    of GTK+ developers.

 B) Not an evident concern of most of our user base. If I want
    get less reports about compilation troubles from GTK+ users,
    first step is providing Red Hat 8.0 RPMS.

I'm not going to promise to be able to spend any time on this 
myself, but if someone wants to work on it, I'll try my best
to provide fast feedback and not to block their work.

> For the time being, we often simply find that our supported platforms are
> simply not considered at all by the Gtk+ maintainers (e.g. Solaris 2.5.1)

I'm not sure what "not considered at all" means. I typically don't
consider *any* platform at all other than the ones I develop
on unless someone submits a bug report or fix.

Solaris 2.5.1 is sort of "special" because it it X11R5 based.
Since X11R6 is almost a teenager now (coming up on its 9th
birthday), I am pretty tempted to declare that GTK+ simply
doesn't work with X11R5.

But the changes are pretty trivial
(http://bugzilla.gnome.org/show_bug.cgi?id=110523 has a patch,
though I'd do it differently), so most likely we'll just fix it
for 2.2.2.

> or that there are critical blocking bugs that requires patches, and
> therefore rebuild of the gtk libraries.

I don't think the interval between a bug fix patch being submitted 
and a release of GTK+ getting out with it has typically been that
big. 2.2.1 => now is probably the longest we've gone without a stable
release since the release of 2.2.0.

> Hopefully this will become less
> and less of an issue now that Gtk+ 2.x is getting more stable.
> 
> > I think your are overestimating here. fonts.conf is really a very
> > simple issue here compared to dealing with the GTK+ and Pango
> > modules.
> 
> I can assure you that building
> 
> expat/FreeType/fontconfig/libXrender/Xft
> 
> on non first class citizen platform can be a real pain (certainly building
> the whole Gtk+ 2.2 suite today is already quite a challenge, so any new
> dependency add to the difficulty).

I'm not denying the pain, I'm just saying I don't think the *extra*
incremental pain is that big.

The basic idea here is that we can provide much better support
on these old platforms if they are using the same code as the
new platforms. Rather than code (PangoX) that is:

 - Untested by any of the core programmers.

 - Very indifferently maintained and just getting worse.

 - A real feat of programming that it ever worked at all
   reasonably (if I do say so myself).

> > It may be irrelevant to your *particular* situation, but it is
> > extremely relevant to your *general* situation, which is:
> >  
> >  I have a commercial application using gtk2 that I want
> >  to install on my customer's old Unix workstations.
> 
> On the other hand, you should be more willing to help the situation of GPL
> software (be it commercial or not) rather than proprietary software.

Perhaps, but I'm even more willing to help the situation of
GPL software and proprietary software at the same time. :-)

Regards,
                                         Owen