updates repo admission checklist hole

Michael Schwendt mschwendt.tmp0701.nospam at arcor.de
Mon Aug 3 16:42:07 UTC 2009

On Mon, 3 Aug 2009 08:14:32 -0700 (PDT), Maria wrote:

> Hi Michel, 
> In my opinion you are wrong because:

Yes, given your sender address, it's a wrong decision to reply to your
strange email at all.

> Fact: update repo has broken packages.

True. The package cannot be installed, and in order to work around it, it
requires users to exclude it (or use --skip-broken) manually. Nobody
claimed otherwise.

> Fact: broken packages keeps sleeping into update repo.

True. Nobody claimed otherwise.

> Fact: QA must consider everything broken until proven contrary.

False. In a project where a lot may be contributed by volunteers, you need
to be careful when to say somebody "must" do something. Further, as I
explained before, there is no global QA in Fedora that has the resources
to handle everything. And there are no automated tests run on updates so
far, either. Ideally, every package at Fedora would be maintained by a
team of volunteers which would include at least one power-user, who
would perform a couple of key run-time tests for sure.

> Fact: You failed to lead situation toward an improvement in Fedora (as mission statement says), because
>          you failed to professionally analyze user comment and make required adjustment.

False. First of all, I'm not occupying any leadership position in the
Fedora Project, and I'm not a Fedora press spokesman either. What I write
is not an official statement made by the Fedora Project.

Secondly, in reply to the synce-hal/synce-serial broken dep, I made sure
that Extras repoclosure detects these kinds of "Obsoletes without
Provides" packaging mistakes in the future. The script, however,
is not run in a place where it would have veto-powers and could block
broken updates. I run it from a remote location. It notifies package
owners, and it's the package owners' responsibility to push updates into
updates-testing first, to give such scripts the chance to be run.

Finally, you can't win anything at all if you attack me from arbitrary
@yahoo.com addresses.

> Someone more open will spot the workaround in a second: 
> "Change the system to give negative karma from start ( or consider zero karma as not acceptable for updates-repo) when a package arrives into updates-testing, thus we will ensure that some human at least tried to install and test the package before it gets with positive karma in updates-repo this will not annoy users" .

Who are you to think that no other people before you have discussed
alternative update policies? Truth is, so far, no relevant committee
has made updates-testing (or +N karma) mandatory. It would be considered
as unneeded bureaucracy by enough package maintainers.

Packagers are free to choose between pushing into updates-testing
or marking updates stable immediately.


> Overall it is bad for fedora public image to have broken upupdate packages. 

True.  For some types of updates, marking them stable immediately,
without giving the Fedora community a chance to contribute testing,
is frowned upon.
> I personally think that you do not reflect the spirit of fedora where people are friendly and open and strive for bleeding-edge improvements.

You mean, let's do more role-playing then? ;)

More information about the test mailing list