On Tue, 2003-02-25 at 09:04, Alfredo Valles wrote:
> Hi people
>
Hi :-)
> I write this email hoping that some of the people who make final
> decisions for the Debian project will read it, that's why I write to
> this list.
>
There aren't really "people who make final decisions", at least not
along the lines of your problems. We're all volunteers, doing this
cause it gives us wood(y) :-) We're somewhere between an anarchy, a
democracy and a flame-fest.
If you have a problem with a piece of software, file a bug report.
That's how we work -- we get your bug reports and they tell us a user
has a problem.
> Have you read the review of Debian 3.0 at
> http://www.debianplanet.org/node.php?id=831
>
> I'm quite new in Debian but I like it a lot, I really do, but been
> honest I find 90% true this review. So I'm worried about Debian's future.
> I believe Debian is the best OS and I'm very proud of use it at work and
> at home. I also believe that it represents the living proof of a new,
> viable, fun and more productive organization of work.
> At least that's what I used to think until I realized how very few
> people are attracted to Debian these days.
>
Debian's "future" ? Are you worried that it's just going to vanish?
The core userbase of Debian (who to be honest are the people we maintain
packages for first) are ourselves. As long as someone maintains a
package, there's at least one person using it (in theory, anyway :).
Debian won't vanish because it lacks market share, we're not a company
out to make money.
It's often voiced that Debian aim for expert users, I don't think that's
necessarily true. As a group we have the aim of a drunk one-legged
spider on ketamine trying to fire a truck-mounted artillery rifle[0].
However it is possibly true that Debian is *most popular* amongst expert
users, because it doesn't try to hold your hand too much.
If Linux distributions were tool boxes, RedHat and co would be a small
clearly labelled and colour-coordinated pocket set of helpful tools for
woodwork class. Debian would be a hardware store.
> I live in a small country, but our linux community is growing fast, we
> are hundreds now, I know many from our discussion forum, yet only 2 or 3
> of us use Debian (or Debian based distros). Everybody loves RH, Suse,
> and Mandrake. Even the most advanced users prefer RH for servers because
> it's the distro they know well and is widely used.
>
Crackers prefer RH on servers too. :-)
*ahem*, sorry that was bad :o)
> I can't recommend Debian to my friends, cause nobody wants to spend 2 or
> 3 days hand tweaking the system, only to get to the point where
> commercial distros leave you 30 min after you put the install CD.
>
I have an opposite experience of Debian. Once installed, I can pick any
random daemon (say MySQL) and have it installed, configured AND working
exactly as I intended with one command:
# apt-get install mysql-server
With RedHat and co, that often involves some pain, especially the
minutiae of configuration.
And as for SuSE, it takes me weeks to get the distro installed from the
CD to a point where it's actually *USABLE*.
Unfortunately our *initial* installation system does suck, because it's
out of date. There is work to change this[1] and the next version of
Debian will come with a brand spanking new installer.
> Personally there are a few things I hate about Woody and I would like
> someone to tell me why these so easy to solve issues are still around.
> For example:
>
> 1-The installation process almost always misses to recognize sound cards
> and network. But ironically, Debian CDs include Kudzu, which have always
> worked fine for me detecting and configuring these devices once you have
> manually installed it with dselect. Isn't that dumb?
> Now that there is knoppix available I don't see any reason for Debian
> not having a real first class installation system with a very good
> hardware database.
>
See above. The new installer will hopefully fix these kind of things.
> 2-After you install the desktop environment with tasksel the gnome fonts
> are "BIG". Took me a day to figure it out how to change that.
>
This is because the default DPI (dots per inch) setting is 100dpi,
instead of 75dpi. No idea why, there's probably a very good reason
though.
File a bug report and see what answer you get.
> 3-All the bash limits are set by default to unlimited except the stack
> limit, which is set to 8 Mbytes. This make programs which work with big
> arrays in memory to die with a laconic "segmentation fault". Took me
> almost a week to discover this one, after I stripped out a large
> scientific program seeking for an "invisible" bug in the code, not
> having a clue it was a system issue. I wonder how many scientists have
> tried Debian for extensive computer calculations and having encounter
> this same problem have simply switched back to another distro.
>
File bug reports.
Altho to be honest, I can't see why any program would have a large array
on the *STACK*!
{
int *a, i;
a = malloc(sizeof(int) * 2048);
for (i = 0; i < 2048; i++)
a[i] = 2 * i;
}
"a" itself is on the stack, using only sizeof(int *) bytes. The "array"
it points to is on the heap.
{
char a[8388608L];
}
That's 8MB on the stack!
Sounds more like a bug in the program to me, 8MB is more than enough
stack for anything I can think of! An unlimited stack would let
infinite-recursions carry away until they took down the system, at least
8MB stops them before they go too far.
Besides, you can always change the limit remember! It's not a
root-enforced limit.
$ ulimit -s unlimited
> 4-Finally but most important: What happens that you haven't included
> kde3 even in unstable until a few days ago? Why so many packages are so
> out of date?
>
KDE is written in C++. The latest release of the GNU C++ Compiler
(g++ 3.2) breaks binary compatiblity with previous releases. Therefore
all C++ libraries are having to be recompiled before the binaries can
be.
KDE needs a LOT of libraries, these things take time.
In contrast, our GNOME packages[2] are reasonably up to date, that's
written in C so doesn't have the broken-ABI problem[3].
If you see another package that is "out of date", submit a "wishlist"
bug report to the package for it to be upgraded. Be nice, the
maintainer's probably either not noticed, or got a very good reason for
not updating just yet.
(This only applies to unstable - stable doesn't get new versions of
packages, we backport bug and security fixes into the old versions).
> I have heard that a bug problem in libc6 is the cause of this delay, but
> I don't see Knoppix or Suse hanging every day, and they do have the last
> soft, so they have found a way to circumvent this problem. Why can't
> Debian do the same?
>
Bugs in libc6 are the reason most packages in testing haven't been
updated for a while. Not the reason for unstable not being updated.
> I have also been told that Debian supports eleven architectures. That's
> really great! But if this prevents from having the last software
> available in an official repository for the arch that 95% of people use,
> then I think something have to be done.
>
Most maintainers build and upload binary packages for i386 straight to
unstable, the autobuilders build the rest later. If it's unstable you
care about, then there isn't a problem.
If testing/stable were getting seriously held up by a myriad of packages
for one architecture[4], I imagine the Release Manager would make a
decision to drop stable support for it. It hasn't happened yet though.
> Mixed would be a kind of mix of a testing debian system with the very
> latest soft compiled on the fly with a system like the one Gentoo has. I
> understand that Debian has some tools to generate a .deb package from
> sources using a custom script, so it shouldn't be so difficult to add
> dependency resolution capabilities to this system, and make this beauty
> available in Debian too.
>
You're describing exactly what happens in unstable for the majority of
other architectures. Maintainers upload an i386 package, the
autobuilders automatically build the package for the rest.
It works, it's what we do.
"Latest software" is decided by the maintainer of that package.
> Being new and inexperienced I have a high probability of been talking
> nonsense with this new schema stuff. But my point is that someone with
> the expertise and good ideas has to come up with something new that
> could help Debian to follow a new paradigm: easy of use, good look and
> latest soft without sacrificing its current strengths.
>
Debian stable's current strength is that it *IS* stable. You can deploy
it on a server, and know that 2 years later when the new stable release
comes along, it's still running.
You can't have latest software *AND* stability.
Well, you can. You can have either "stable" or "unstable".
If you want to try something in the middle, and don't care about the
security of the machine, we have "testing" as well. Great for desktops
-- you get not the bleeding, but certainly the grazing-edge of
software[5], without most of the death and destruction unstable can
cause.
Of course, using testing means *YOU*, rather than the Debian security
team, are responsible for finding and fixing security updates. Often
taking the newer version of the package from unstable is sufficient when
you see a DSA.
> One would get the impression that Debian is now a bit bureaucratic and
> reluctant to changes. Maybe it is not, but that's how it feels like for
> a newcomer. I wonder how decision making is conducted in Debian? Is
> there a guy (or a closed group) who take decisions like the holding back
> of so many new soft, because of the libc6 issue? How can we (the users)
> participate in the decisions of really important matters? Is Debian a
> democracy? (I hope so)
>
Package decisions are made by the package maintainer. If you've a
problem with a package, file a bug. If you think they've done something
wrong, be polite and persuasive.
There are many groups internal to Debian, there's a group working on the
new installer, there's a group who do QA, there's a group who vet new
maintainer applications, there's a group who maintain the package
systems. Lots and lots of groups.
And every discussion they have is OPEN, and held on OPEN mailing lists.
If you want to help, join the mailing list and offer assistance testing
things.
For archives of all discussions ever, see: http://lists.debian.org/
If you want to join the project and maintain packages yourself, you can
too. See: http://www.debian.org/devel/join/
Now try asking RedHat, Mandrake or SuSE the same questions. I'd be
interested to hear the answers :).
As for who's made the decision that libc6 (and most packages) are held
back from migrating from unstable (where they are the latest versions)
to testing - that decision was made by a computer program designed to
keep release-critical bugs out of testing.
Scott
[0] Slight Blackadder moment there.
[1] d-i: http://people.debian.org/~mbc/di.html
[2] <flamewar> Which is better than KDE anyway :-) </flamewar>
[3] <flamewar> Which is better than C++ anyway :-) </flamewar>
[4] libc bugs don't count
[5] Unless libc has bugs of course
--
Scott James Remnant Have you ever, ever felt like this? Had strange
http://netsplit.com/ things happen? Are you going round the twist?
Attachment:
signature.asc
Description: This is a digitally signed message part