Re: [desktop] Unix configuration nightmare
On Tue, 2002-10-22 at 04:05, Mark Howard wrote:
> Hi,
> There has been much debate about unix system configuration [0]. This
> has resulted in discussion and even a message of support from the fsf,
> however no actual work has been done.
Before we get into details, let me be very blunt. I do not think this
particular sub-project is going succeed, or that it even really makes
sense.
I know I said earlier that I'm a strong believer in the power of free
software, and I am, but what you are talking about would require an
enormous investment in time from the community. The implementation work
*alone*, to cover even 60% of what's out there, would take more
person-hours than I think the entire Debian Developer community has to
offer for the next several years.
And not only that, you would need at least a full-time team of 30-50
people to handle the resulting flamewars that are sure to develop when
you try to replace someone's idea of what a configuration file looks
like with something else, be it XML, Bind format, Windows INI, or even
something pluggable like GConf.
And finally, I think the entire idea is on somewhat shaky ground. I
will elaborate further on this below.
> - Write an entirely new configuration format to be used by all programs.
> Write friendly programs for editing these
Again, seriously, "all programs"? Think about apt alone. Think about
how much time it would take to sit down and write (for example) an XML
Schema for /etc/apt/preferences, /etc/apt/sources.list,
/etc/apt/apt.conf. Then think about how much time it would take to
actually convince Culus this is a good idea. Then think about the
enormous work in implementing it, documenting it (!), and debugging it,
and finally: maintaining it.
And that's just ONE program.
> - Modify/Merge webmin/debconf/xst/gconf/... s
Ok. Here is where I think you are somewhat misguided. Debconf and
GConf are *very* different, in some both subtle and overt ways. For
example, Debconf is very tied to the system-level configuration. It is
designed for configuring a program once, when it's installed. It really
has no notion of per-user preferences.
GConf on the other hand, is designed to be run on a per-user basis, and
is for storing changing user preferences over time. This has affected a
great number of things in its architecure. For example, when you change
an entry in GConf, interested applications are notified of the change.
That's why very very few GNOME preferences dialogs have an "Apply"
button anymore. It's just not needed; changes take effect instantly
(unless they're dangerous). Debconf has no notion of anything even
*like* this.
You can take the GConf notification away from me when you pry it out of
my cold, dead hands :)
> The problem: something has to be done for every program being used.
> Therefore nothing concrete has been done. If that effort was split
> between say, 1000 developers, it would be far easier.
I doubt that even 1000 developers (which we are not even close to
having, I think we have about 250-300 active) would not be enough to
convert even half of Debian's source packages (say about 3000) in any
reasonable amount of time.
> - do you think this should be a (long term) aim of debian-desktop?
In short: No. I think the status quo is generally speaking,
acceptable. Debconf serves as a system-level "wizard" for program
installations. You should think of it a lot like InstallShield in
Windows. GConf takes the place of the Windows Registry (although it is
far, far cooler than the registry). And stuff like Webmin and the GNOME
Setup Tools take the place of the Windows Control Panel. As far as I
see it, it simply does not make sense to attempt to unify these things.
Now, a smaller goal, like adding GConf support to programs where it
makes sense, is a far far more reasonable goal,and one I would actively
support.
But in the end, you could prove me wrong. If you still, after reading
all of this, think what you're talking about is a good idea, then there
is nothing stopping you. That's the great thing about free software.
Reply to: