30 Nov 2000 (updated 2 Dec 2000 at 07:26 UTC)
»
Well, what the hell. Let's try something new and write a
diary entry.
Since I'm not worthy of posting replies to the
articles
(Probably with good reason), I'll reflect on my own
meandering experience here instead.
The modest
proposal certainly sparked a good amount of discussion,
the vast majority of which is of good quality, as
lilo observed in a comment
towards the
bottom (at the time of writing. 'The bottom' was comment
#29') A refreshing improvement over the usual Slashdot
fare, to be sure. My
initial thoughts after reading
RyanMuldoon's article
were agreeable as far
as GLib, more or less agreeable regarding gconf and Bonobo,
but uncertain where gnome-vfs is concerned.
The benefits of integrating all of the above
libraries
are
plain, it's simply a matter of if those benefits justify
introducing the libraries into a standard GNU environment.
With GLib, this is pretty simple. It's a useful library
that adds valuable features to any libc environment, so long
as it is being put to use. Obviously any system running the
Gnome desktop is using it, but at present the use of
GLib, afaik, ends there. Would introducing Glib into all
GNU environments as a standard library expand that use?
Perhaps.
As I see it, there are two ways a library
becomes
'standard'. The environment declares "This is standard," or
the people declare "We want this to be the standard."
Which method is responsible for more standard libraries, or
which method yields more successful standards? Personally I
would lean towards the latter. If GLib truly belongs in a
standard environment, that should become evident naturally
as programs beyond the Gnome sphere of influence make use of
it.
My familiarity with gconf is virtually
nonexistant,
so I
won't spend too much time blowing smoke. Unix, in its long
life, has acquired a rather substantial inertia in many
respects to how it works, not the least of which is how
configuration is done. This exact issue was discussed at
length some time ago on debian-devel, when the suggestion
was brought up to move all Debian config files to an XML
structure (IIRC, gconf hadn't been written at the time).
The idea raises a series of important questions. How do you
do that at a distro level without the upstream maintainers
adopting the structure? Would they adopt such a structure?
Would it become a peculiar feature limited to Debian
packages, forcing Debian package maintainers to work that
much more? And so on. So, as we ponder gconf, would a
'standard' configuration library be accepted by application
writers beyond Gnome's influence? As with GLib (Moreso,
actually) I think the best way to determine this is to see
if it's being used, rather than push its adoption. I can
already hear an argument against this. "But if people
aren't aware of gconf or its benefits, how would they know
to use it?" Well, that's an education issue. Get the word
out. Tutorials, presentations at conventions, articles, and
all that good
stuff.
And speaking of good stuff, we come to Bonobo.
Iain's pipes-for-GUIs comments
were right
on, I'd say. What I wonder about is if standardizing it in
the GNU environment is even necessary, since its usefuless
is focused on
GUI-type apps. I think wider adoption of Bonobo would serve
to benefit applications, though I wonder about apps that,
for whatever reason, find themselves wanting to communicate
with KDE's component system and Gnome's at the same time.
For some reason that seems like it could get hairy. Anyway,
I think Bonobo can stand, and thrive, on its own merits
without being declared a standard library.
This leaves gnome-vfs. Whatever fate it may
bring
me, I
can
at least understand aaronl's
reservations
about seeing this sort of utility enter a standard
environment, but my reasons are somewhat different. Where
he sees unnecessary 'over thinking' (for lack of a better
term) on the part of the Unix system, I simply see overhead.
This is possibly because I'm quite fond of
Debian's almost anal habit of breaking packages
down into their smallest logical components. This has, for
the last five years, consistently
yielded a more efficient system that occupies, on average,
half as much space as a comporable RedHat-based
installation. I like that. So when I imagine a
(potentially large) vfs being tied to my system at a
relatively low level, I can't help but worry. As
egnor quickly pointed out, wget
| wc -w
works just fine. :) Would it be cool to grep a web page
without a pipe? Sure. I could think of countless uses for
my standard Unix commands to take URLs as parameters, but is
it the right thing to do? Well, that's certainly not my
call.
Back to work.