Re: Debian desktop -situation, proposals for discussion and change. Users point of view.
Hi, Greg
You took it quite actively.
As many see, all of them are different in server and in desktop world,
and many times Debian chooses to dictate the users "we know the best
what You need" instead of listening to them.
Why then are there 28000+ packages in Debian? If Debian only dictates,
why then are there *FAR* more packages available for install than in
*ANY* other Distribution? How many Window Managers? How many alternative
packages to do the same thing, like word processing, editors, music
clients, rss feed readers, web-browsers? I could go on for days, but I
hope you get my point.
Come on, we know the answer, you can say it.
Yes, no single other distro offers such a vast choice possibility, if
we're speaking about software. The "dictate" I feel on other levels.
Diff the end user approach of Ubuntu and end user approach of Debian and
You see a part of it.
It's complex to discuss undercover, however with Ubuntu, user get's an
_impression_ that this is created especially for _him_ and that Ubuntu
_cares_ about what he might need. We could call it "marketing", however
it's only partially about marketing. Whatever quality the Debian offers,
it's harder for user to _interact_ with the community, and harder to get
the impression that he actually can have any impact on what's going on.
One easily gets impression, that he can move the mountain more easily
than affect Debian's course.
b, Stable without (too many) crashes
Do you realize Debian's stable is classified as this:
Stable means stable package list. No changes in API and ABI
names or versions. This means no newer versions will ever make
it into "stable". It is in "maintenance mode". This makes a very
good setup for those wishing for "Rock Solid" machines. Doesn't
crash. "too many" comes from the "Windows World", does not
typically apply to Debian's Linux.
No changes, no newer versions => dosen't crash? It's simply not true.
For example, the Debian Woody used an ancient version of Mozilla. _Very_
crashy one, compared with newer versions that came few months later.
Noone could call that "stable" one.
Generally speaking, there _are_ stability issues in any software. Should
they eventually get fixed upstream, then newer version _objectively_ is
_more_stable_ than older, providing no new stability bug has been
introduced since the old has been fixed. Yes, it's perfectly possible
that newer version of software is more stable (less crashy) than old
one. (Should it be reversely, then software is more and more crashy and
will not be usefull at the end ;-)
As I said, "old" is not automatically equal to "stable".
c, Applications should work generally
Okay, what specifically does not work in Debian?
I just listed criteria, didn't blame Debian at this point.
d, Applications should work together well
Again, if you are using a Desktop environment, they just DO.
By the means of usability, not always.
For example, Abiword dosen't exchange files ideally with other office
suits (Koffice, OpenOffice.org etc) found in Sarge due to different
import/export filters. With Etch, it's been improved (due to upstream's
work, of course). However, they and other apps are being under
development that leads to ODF support. New version will work
_much_better_ with each other.
Openoffice.org hve had problems with importing it's own files, that have
been fixed. Thus newer version is more interoperable with itself than older.
Other example is SVG support. We'll (hopefully) get soon new version of
OpenOffice.org with SVG support, Firefox with improved SVG support, etc.
Applications mature in course of interoperability in FOSS world. "Newer"
almost always meens "better".
In fact, I use XFCE. If I click on a link in my e-mail client
(Evolution) it opens up my preferred Web-browser (Iceweasel). If I open
a "Word Document" in Iceweasel, it opens the doc in OpenOffice.org
writer. If I make a mailto link in Writer and click on it, it opens an
Evolution "new mail" interface. So, once again, I don't see your problem
here.
Well, if You have chosen to use Thunderbird (Icedove) instead of
Evolution, You must have installed gnome-support manually, otherwise it
dosen't interact with other apps well.
In Sarge, I've had many problems regarding file associations with
Thunderbird.
I just say, that newer versions usually interact better with each other,
and thus the oldness is decreasing the usability, not increasing, by
means of interoperability.
e, The serious security problems should get fixed ASAP
Again, just pointing the need, not blaming anyone.
Debian's Stable cannot introduce new versions. This complicates things.
It makes it tough, the security team has to "backport" the fixes from
the "new versions" and force the changes to not bump the ABI numbers.
This may seem trivial to you, but it is NOT.
In fact, Im saying that it is too complicated (if even possible) to put
new patch to old software, than follow upstream.
Have you even looked at Debian's BTS? Unless your patch or bug is a
security related problem, it won't make it into "Stable". "Stable" is in
maintenance mode.
Yes, that's the problem I'm speaking about. No chance to usefully report
usability issue in stable and see it fixed. Upstream dosen't solve
ancient version's issues.
g, I _need_ the new features of some applications -for example
improved import/export filters and so on, and I need them now, because
yesterday it has been already late
Then compile it and install it yourself. If you need something bleeding
edge, then you will have to do it yourself. *NO* Linux distribution does
this.
??? I'm speaking about reasonably recent stable upstream versions, not
nightly builds.
Ubuntu comes with recent software. Debian offers old software from the
first day of stable's life cycle.
I do it everyday. My youngest daughter takes LiveCD from the Debian Live
project to school, everyday. She typically hands out 2-3 a day. The
school system contacted me about this possibly pirated "Linux" my
daughter was handing out as all the kids wanted to run it at home.
The threatened to report her to the Local Police and things like the BSA
(which is like the RIAA and MPAA, but for software). I said, please go
*READ* the licensing of the Debian LiveCDs
They called back and apologized, asked if they could keep the CDs they
confiscated from my daughter. I asked if they wanted more.
Very nice. I'd like to read more such stories.
We must make clear that:
1, Any distro is only as good as the software it offers.
So, with 28000+ packages in the Debian Repositories, what does this say
then?
Be aware, quantity dosen't implict quality :-) Yes, managing that amount
of software is colosal work.
2, Any software does, and will have, bugs.
Of course, many have yet to be discovered.
Exactly! Even that hidden in "rock stable" versions of software in
Debian stable's repositories!
And the user should be able to report them and see them fixed.
New exotic or *EXPENSIVE* hardware doesn't make it into the kernel for
years sometimes. Mainly because of the "Windows cover 90%+ of the markey
why bother." attitude many/most hardware vendors have.
Cannot be answered just that way. I don't use new, exotic, expensive or
whatever hardware. I'm quite conservative. However, I couldn't put Sarge
out-of-the-box on ordinary hardware because the 2.6.8 kernel version
didn't support it well. The same with old HP nx9105 notebook. And so on.
I compiled kernel myself and whoila, worked well.
I'm sure I'm not the ony one with such experience. That puts a question
in here: is _that_ the way it ought to be?
Should the distributional kernel be just an artefact that often dosen't
even allow flawless installation (sometimes the only way is to install
on another machine, compile kernel and THEN put it to the target one..)?
Or should it be improved somehow in Debian, for example, offering newer
stable kernel version sooner than with next release?
Please give us example of these "stability" issues with applications.
Anything you care to REALLY share with us?
I've already had. And there are more. Just look at the GNOME
applications in Sarge and You have lot of inspiration about stability
problems with "stable" software nailed in stable Debian. Look at Totem
in Etch.
If You really want, You will certainly find them.
But, security in Debian *IS* addressed in most cases. Sometimes when a
security exploit is in a common software installed on desktop machines,
these backports of the fixes from newer versions for exceptionally
complex software may take a while so as not to bump any API/ABIs in any
way. Barring being able to backport the fix without further compromising
the software, it may even be removed from the next revision of "stable".
Yes, EGroupware has been kicked out by similar means; maintainer was
unvilling to let the security fixed upstream version in, and unable to
fix the old version himself.
However, the security upgrades repeatedly caused software to stop
working well in Debian, so the software version's rigidity dosen't
really help much. It simply dosen't prevent software from breaking.
Security upgrade repeatedly? No. you are thinking of another distro.
Haven't heard how libtruetype security upgrade caused OpenOffice.org,
Wine and other apps crashing for 2 months until got fixed?
Never had problems with security patched slapd?
You're probably lucky one.
Fedora and SuSE and Mandriva just include the latest software in the
release, which changes the ABIs abd API that Desktop Environments are
expecting to use and see. That breaks things. Not they way Debian does
security releases.
That's the teory that is broadly believed in for years.
Do you even subscribe to Debian Security Announce?
I watch them.
Current stable upstream versions of any software should not have major
usability issues anyhow.
Current When?
Current, recent, upstream released, call it how You wish.
However, if there are major usability issues in software in Debian,
should they have been fixed upstream, user must wait for next stable
anyhow to see the fix.
Usability issue, you keep using that word as a magic statement.
I'm speaking a bug that affects the user's experience with the program.
Program should do A, however, does B, or crunches data, or... Simply a
bug. I call it "usability issue" to differ it from "security issue".
Debian
*IS* usable the instant it is released. It may have some errata. Does
Microsoft publish those things EXACTLY at release of a new Version of
Windows? NO. Not at all. Wait. How long was it between Windows XP and
Windows Vista?
I don't watch M$' approach, and don't compare it with Debian. I confront
the Debian's way with desktop user's needs.
Any security patch can affect the usability of software, either by
backpatching an ancient "stable" version, or by installing the new
fixed upstream version. I personally would prefer upstream fixed
version, before the hacked-around-patch that tries to fit the ancient
piece of software that the patch has not been created for originally.
Okay, so you are saying that "suppose" apache v2.2 have a severe
security exploit that forces the ABI to be bumped. Are you telling me
that you would rather Debian just compile the new version and publish
it?
I'm speaking of _DESKTOP_ software.
Please, consider the criteria to be a bit different with server. I don't
think much desktop apps are such dependent on each other that something
would seriously break if some of them got upgraded.
MOST Debian desktop users I know of, are using "testing". This fixes you
problems in great instances of "just" use testing.
Testing has been discussed in mails before.
Installing TESTING or UNSTABLE package on STABLE does typically RESULT
IN A BROKEN MACHINE. There is a reason for that... the "specialized
versoin" of libraries you complain about. If you don't understand and
don't want to learn about it... well, I am feel sorry for you.
You haven't read carefully.
I have named the facts. That are the reasons, why user cannot get a fix
for his usability issue or new package version in official, advisable way.
All in all, for the user, reporting the usability bug is waste of
time, and if he tries to help fixing it, he may end up with
reinstallation of the system.
I have installed Debian on my machines exactly ONCE. Actually less the
ONCE per machine. I just scp the installation over and change things for
the new machine. Done. You obviously do not get the Debain System of
things.
I'm not apt-insane, so if I attempted to install unstable package into
stable, I certainly _would_ end up with reinstallation.
For the upstream, the Debian user is out of the game and out of the
community.
No it is not, the Debian user needs to use the Debian BTS. reportbug
easily allows users to report bugs on packages. I've used it recently
even, and have had great results with it.
I'm glad it has helped You.
-Debian user is facing the bugs that usually have already been fixed
upstream ages ago.
Stable is Stable. How many times must I go through this.
Yes :-)
How many times must the desktop user go thru this, on the other hand ;-)
This is damaging the ecosystem on both sides.
No, it is not. Why do you think SO MANY DISTRIBUTION ARE BASED ON
DEBIAN?
Speaking about desktop, I know just Ubuntu, that uses recent versions of
software.
g, Need for new features
You are out of luck with Debian. The software in Stable is already
out-of-date at the day of release, quickly fading to museal
experience.
backports for Debian stable (others have listed it), look at it, embrace
it. Love it. Don't assume. This though, does NOT make stable == STABLE,
as these new features are not as vetted and reviewed by as many. This
also is not officially supported by Debian though, as it is NOT stable.
That's a problem.
If You need new software, You must compile it Yourself, tackle with
generic installers, risk the system's consistency by adding testing
repositories, or when You have enough of that, choose other distro.
Ha, you really are calling the pot, dayglo orange here. Debian Polcy
forbids Stable to have features added. Again Stable == Stable.
Yes, The Policy. Did I say at the beginning, that _desktop_ user dosen't
get an impression that Debian cares about his _desktop_ needs? ;-)
The Linux Desktop in Debian is not its primary function... won't be
ever.
You see it. So, why don't consider the desktop to be a special part of
"universal operating system"? Why don't make some changes to give the
desktop more of that it needs?
The release cycle should be completely redesigned.
Completely redesigned? No. Tweaked? yes.
Okay.
The desktop-oriented packages should be available in as recent version
as possible. The desktop software must not trespass the whole
unstable/testing/stable scrutiny. User should simply get the latest
stable upstream version, no more, no less.
Why? If you MUST have bleeding edge, run Sid (unstable, which means
unstable packaging and APIs and ABIs, not instability like most people
assume).
Installability, dependencies.. Afraid of that.
The distribution cannot, and should not attemp to, isolate user from
software bugs. There always ARE and WILL BE bugs, even in the software
that made it into "stable" Debian, even in "security-only" fixes etc.
Managing the bug reports for old versions of software and backporting
the fixes is waste of time.
Do you understand *WHY* these things are backported? You obviously have
not considered the whole "recompiling of all redepends packages" to get
them to work with a "new upstream" package
With desktop software, I don't know... Please, name some crucial desktop
app that couldn't be upgraded without need for deep dependencies
recompilations.
The stabilisation of software is UPSTREAM's business. Let the upstream
do its job. Let's give the upstream more responsibility together with
support.
No, stabilization IS NOT UPSTREAMS BUSINESS. The Linux Kernel in
particular. Firefox as another, Apache yet another... etc. Also getting
everything to work with one another. You wouldn't like Apache not be
able to talk with postgresql due to a typo on Apache or postgresql
fault. Or what about Evolution not running properly in KDE. KDE doesn't
care about evolution, they have kmail. Come on pull you head out.
Hm, I don't use Evolution, nor do I KDE. I didn't know they don't work
together well.
Let's not "over-stabilize" ... So you want it to be Windows? Just Code
It, Compile It and push it out the door?
If it is stabilised upstream, well, yes :-)
Using 2+ years old desktop software, that's what I call
"over-stabilisation".
The thorough stabilisation scrutiny should apply only to server
applications and the basic platform, with the optimal length of
release cycle of maximum 18 months. During stabilisation, LSB
compliant basic platform (libc and other libraries) should get
established and polished to high merits.
So, now it sound like you are suggesting another "distribution" called
"Desktop".
New repository with desktop oriented upgrades.
All, I can say is this: You do not understand Debian and its core values
and policy and volunteering. Debian *IS* after all a Volunteer
organization.
Maybe. However, if I don't understand, after fighting with 3 releases of
Debian for useful desktop that fits my needs, then please don't be
surprised if newbie would just see, turn around and go away.
There is NO lost community. The whole reason Debian HAS a Bug Tracking
System is to aid it in making a stable release, doing give back to
upstream, doing innovation that upstream refuses to even consider,
unless they see it working. (Think exim for a minute on that)
I don't know exim. However, if it worked just half You have written,
that's great.
I work with ordinary mainstream desktop applications, and don't see it
so colourful as You do.
Should there be really bad bug in upstream's stable version, let's
give the user a tool for simple downgrade.
dpkg does this by default. Ding, done.
Simple? For desktop end user? No way.
Okay, so when you install reportbug, doesn't report EVERYTHING on a
debian system? Yes.
Yes, and developer asks, whether the bug is already reported upstream. I
saw that. And whether it hasn't already been fixed in unstable. Would
ordinary desktop user install packages from unstable? In last day of
Debian use, maybe.
Moreover, on top of that, every distribution has its own bug tracking
system. This is not bearable. The user should be able to report a bug
from central application, without need to register here and there. One
registration, the distributional one, should be enough. The
application's developers should work with issues on unified way, not
searching reports and information in every-distro's mailing lists etc.
Searching and searching and searching... The distro you are using... if
it is worth ANYTHING has a great bug search system. Has a GREAT mailing
lists search tool. Or you have Google. Learn to use Google.
I'm registered in two dozens of great bug search systems and I'm already
sick of that.
Bugs are found routinely. Let the user to report them routinely too.
Save him from the whole today's bugtracking nightmare.
*3,
New official desktop repository
In order to give the users painless access to the fresh desktop
applications, new repository should be built and officially
supported.
So, who owns it or maintains it?
That's the question.
Security upgrades should simply go thru this repository by naturally
upgrading the whole software packages to recent upstream version -no
longer backporting the patches.
So, Apache falls into this? (an other server type things)
Yes, it does, providing Apache is _desktop_ application.
(Of course it dosen't).
The repositories should be advised during installation and user should
easily choose, whether he wants to upgrade the desktop software or
not.
Oh, like they currently are able to choose a on-standard mirror already>
or a Standard one? Or combos of them all?
Yes, just like that. "Do You wish security upgrades?" In fact, we no
longer ask that question already. "Do You wish to get latest software
upgrades? Please note, that newer versions of software can change in
behaviour, features and so on. If not sure, answer No".
Debian has never TRULY been a "Desktop" platform, though I have been
using it as "Desktop" platform for years. I don't see it as any less of
a Desktop now as I did in 1999. In fact, it is MUCH easier to use as a
Desktop now than in 1999.
Yes, my impression is exactly the same as Yours.
Friendly
Peter
Reply to: