Re: Document reasonable portability targets

[Bruno Haible]

Paul Eggert wrote on 2011-01-29 in
<http://lists.gnu.org/archive/html/bug-gnulib/2011-01/msg00545.html>:
> The change to gnulib.texi looks good, but my kneejerk reaction to the
> proposed change to gnulib-intro.texi is that although much of what's
> proposed is useful, it divides software into categories pretty strictly
> and this strictness might cause confusion and problems.
> ...  Overally I suspect it'd be better to keep support levels a little
> fuzzy, and not to try to define terms like "essential" and "minor".

OK, I can simplify these categories by indicating how often we test on
these platforms: frequently, occasionally, rarely.

> Instead, how about something like the following:

The question your text is answering is: "As a Gnulib developer,
which priorities should I have?" My text answered the question
"As a Gnulib user, what amount of portability can I expect?"

I think everyone is free to set his priorities himself; there's no point
in writing this down here.

> Gnulib works on a number of platforms that we call the "reasonable
> portability targets".
> 
> GNU platforms, such as glibc, have the highest priority.

This is not universally true. We invest more in Solaris portability
than in GNU/Hurd porting.

> Next come 
> other free-software platforms, such as Cygwin and FreeBSD.  Then come
> proprietary platforms that fit well in the Unix/POSIX tradition

It's not really the free vs. non-free here. We put more effort into the mingw
port than into NetBSD and Haiku ports - although NetBSD and Haiku are free.
And whether Cygwin counts as a "free software platform" or not - I don't
want to discuss this; it would be a fight about words.

> as MacOS X and Solaris.  Then other proprietary platforms that are a bit
> of a stretch, such as mingw.  And last comes proprietary platforms
> that would be so much of a distraction to support that Gnulib
> deliberately does not support them, such as MS-DOS.

This paragraph does not really say why MSDOS is a "distraction" whereas
mingw isn't.

The real criteria are the number of users of these platforms and the
amount of cooperation we get from their supporters. I think this should
be mentioned.

> However,
> already-existing Gnulib code for now-obsolete platform versions is
> typically left in place unless it would significantly impede
> maintenance on modern platforms.

Good point, this ought to be mentioned.

> The exact set of platforms and platform versions, and their level of
> support, is open to judgment and depends on how much work developers
> are willing and able to contribute.  Volunteers to help support
> other platforms are welcome, but should keep in mind that Gnulib's
> goal is to support production applications, not computer museum
> pieces or software research projects.

Yes, good point.

Bruno
-- 
In memoriam The victims of the Zaklopača massacre 
<http://greatersurbiton.wordpress.com/2010/02/12/interview-with-nihada-hodzic-survivor-of-the-zaklopaca-massacre/>