[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: How to make Debian more attractive for users

On Fri, 23 Jul 2010 01:59:53 +0200
"Jesús M. Navarro" <jesus.navarro@undominio.net> wrote:

> Hi, Neil:
> On Thursday 22 July 2010 20:28:49 Neil Williams wrote:
> > On Thu, 22 Jul 2010 10:53:53 -0700
> [..]
> > Removing packages from testing does not remove them from any existing
> > installation, so it's hard to see how the removal of packages which are
> > plainly not suitable for release in stable supports an assertion that
> > testing is somehow not intended for real users.
> Having a system with packages "which are plainly not suitable for release in 
> stable" doesn't ring a bell in your book?

So what is Debian expected to do? Leave the package in testing whilst
the bug is tested and fixed in unstable? You can't have it both ways.
Packages in testing will have bugs reported against them, some of those
bug reports are release-critical but not all release-critical bugs
affect all users equally. If the bug cannot be fixed, there is no
alternative but to remove the package from testing and unstable.

Testing is not 'stable'. As the name suggests, it exists for testing
purposes and testing brings new bugs to light or it isn't worth the
name. If you want stable software, use stable but real users still need
to use testing so that some real user testing does occur. The results
of that testing need to be taken into account, otherwise poor quality
software ends up in stable, which nobody wants to see happening.

Debian Testing will contain packages that have release-critical bugs
from time to time and some of those bugs force the removal of that
package from testing because there's no other way of fixing the issue.

Critical bugs cannot be ignored and buggy packages cannot be left in the
archive to trip up someone else - however, those who make a conscious
decision to keep the package can continue to use it until some other
change elsewhere means that the package can no longer operate. (Obsolete
packages like that will just collect more and more bugs because nobody
is working to keep the package in step with the dependencies etc.)

There is no magic wand to fix all critical bugs in all packages (let
alone all important bugs) and bugs will appear in packages in testing or
there is no point having 'testing'. There is no point just bleating
that the bugs all need to be fixed either - it simply isn't possible.
Some bugs just have to be fixed by removing the package. Better to
remove it from testing than leave the buggy package go into stable.

Debian work is voluntary. If you've got ideas on how some RC bugs can
be fixed, post on the relevant bug reports. That's the only way to stop
a buggy package from being removed from testing - stay alert to the
*results* of testing being a test environment and find someone / help
someone who can fix the bugs that are revealed. If you care about the
package, get involved.

If you cannot do that, then either don't use testing or don't complain
when someone else fixes the problem by removing the package.

> > There are no "internal release master reasons" - there are Release
> > Critical bugs
> How do you think a bug gains "Critical" status?  Is that the kind of software 
> you'd want installed in your system?

If the specific bug doesn't affect me directly, then why not? A
security bug doesn't affect me unless the machine running the package
is exposed to the wider network. A FTBFS bug doesn't affect me unless I
need to build that package. This is why we have a script called
rc-alert - it raises these issues with the local admin and lets the
local admin decide whether the package should be uninstalled or whether
the local admin might even have some ideas on how the bug can be fixed.

It isn't that hard to run:

$ sudo apt-get install devscripts
$ rc-alert

> > and if anyone in Debian feels that the RC bug which 
> > caused the removal of the package was invalid or not as bad as
> > reported, then that person needs to get involved and disprove the bug
> > or explain why the severity should be downgraded. If users don't do
> > that, there can hardly be complaints if those publicly discussed issues
> > cause the removal of the package from Debian mirrors.
> And once a package is removed from Debian mirrors because it is in so bad 
> state even Debian Developers can't stand allowing it being installed on third 
> party systems, how exactly does it become uninstalled from all those systems 
> that unawaringly did installed it?

The admin runs rc-alert and makes their own decision. 

Debian does not use the Kindle distribution model.


Neil Williams

Attachment: pgpVu64t5OAyq.pgp
Description: PGP signature

Reply to: