Re: Did GNOME go 1.0 too early?

[Warren Young <tangent cyberport com>]

Andy Kahn wrote:
> 
> On Wed, Mar 03, 1999 at 11:47:31PM -0500, Havoc Pennington wrote:
> > On Wed, 3 Mar 1999, Andy Kahn wrote:
>
> so until the whole Unix world goes with a popular Linux distribution
> that has a package manager, or unless GNOME releases start providing
> precompiled packages for each popular Unix (e.g., Solaris uses
> 'pkgadd', Irix uses 'inst', Digital Unix uses 'setld', etc.), the
> tar ball installation from source just doesn't work.

Actually, the README file (you did read it, yes?) in the main GNOME
directory documents the installation order quite clearly.  When I
compile GNOME, I just open the README file in a separate xterm, so I can
see it from the one I'm working on and start compiling.

Especially in recent versions of GNOME, you no longer have to go into
code-hacking mode to get it to work.  Generally, for example, a bare
./configure does the job without any fuss.

As for the RPMs versus tarball debate, IMHO you should not use tarballs
unless you really know what you're doing.  (This doesn't apply to
Slackware users, of course.)  I just installed 0.99.8 or so on a box at
work from RPMs, and the hardest part was finding the older Guile
libraries it needed.  (I'd since blasted the 1.2 that comes with RedHat
with 1.3, so I had to revert.)

(Just for grins, I'll register my opinion of RPMs right now: I'm no big
fan of them except on "stable" boxes that don't see much new code.  My
box at home tends toward fewer and fewer installed RPMs because I have
to keep uninstalling old ones and compiling newer versions of the
packages from source.)

That brings up my only real problem with GNOME: it depends on a lot of
outside libraries that are evolving quickly, and it always seems to need
the most recent version of each.  I sometimes feel guilty for all the
traffic I put on ftp.gtk.org by downloading gtk+ and glib over and over
and over just to get the next point release because some silly tool
needs it.  ...And then there's the existing RPMs that the new libraries
break....  This isn't about Gtk+, though because 1.2 will fix this.  My
complaint is more general than that.  Guile, imlib, ORBit, etc. are all
works in progress.

I'm not really complaining, because I think the root problem is simple
code immaturity.  Once things begin to settle down, no doubt a "popular"
version of each library will come out and most people will use it,
allowing newer versions to take their time coming online.  I think
perhaps one of the reasons for GNOME 1.0 is that this is already
happening: Gtk+ 1.2, Freetype 1.2, Guile 1.3, ORBit 0.4 etc. are all out
now.  I think we've just witnessed a mass stabilizing of a whole lot of
code, and GNOME is just one element of that event.  (Not stable as in no
bugs, stable as in "workable enough to allow less frenetic development
efforts around future versions".)

Consider Perl: I don't know what its early history was like, but I
suspect that Perls 1.0 through 3.x were a whirlwind of patches and
releases.  For the last several years, the only "serious" versions have
been 4.0.36 and 5.0.  Sure, each got patched a few times, but in
general, all existing Perl code works with one of the two, and usually
both.  The same goes for Tcl/Tk, Samba, Fvwm, etc.  These packages are
all stable now, and I think GNOME will follow in their footsteps, but it
just can't happen right now.  We wouldn't want it to happen.

I guess the moral of the story is, we're on the bleeding edge here, so I
don't think we have much cause to complain.  Sure, we need a constant
mild unease to keep us going forward, but let's not forget that "point
zero" is for early adopters.  Sure, GNOME may get a blast of "blah"
reviews, but then Windows 1.0 didn't do so well, either.  

Keep in mind, in some ways we're actually ahead of Windows and KDE, and
this is only 1.0.  Surely 1.1 and 1.2 will really rock, and 2.0 will be
unquestionably cool stuff.  Can Windows say this?  I think the bugs and
problems everyone is seeing are really small potatoes, because we're
really quite far along.
-- 
= Warren -- http://www.cyberport.com/~tangent/
= ICBM Address: 36.8274040 N, 108.0204086 W, alt. 1714m
= "I think not," said Descartes, and promptly disappeared.