FESCo wants to ban direct stable pushes in Bodhi (urgent call forfeedback)

[Patrick MONNERAT]

Kevin Kofler wrote:

> at the FESCo meeting on Tuesday, everyone except me seemed to be set
on wanting to disable the possibility to queue updates directly to
stable in Bodhi.

As you say, there are quite a lot of situations where direct stable
update is needed.
This proposal is probably inspired by a very few bad cases, like the
recent dnssec problem (among rare others).
Although disabling the feature will resolve such problems, it will bring
others, much more frequent:
- People test new packages only if they are involved with: they are bug
reporters, they are aware of a particular bug or a new package version,
or they watch the package because they wait for some precise fix.
- Normal users and most packagers NEVER test packages: they don't even
know their existence.
- People effectively testing packages and commenting bodhi for karma
bumping are very few.
- Many BZ reports are a 2-people conversation: the reporter and the
packager: even in the case the reporter positively comments testing in
bodhi, this is not enough to exhaust the default karma count.
- In the dnssec case mentionned above (sorry: this is an example, not an
attack!), the faulty package might have been promoted to stable without
testing even if stable push had been disabled. In such a case, how long
will it take to have a real fix for everyone ? Much longer than the 2
effective days needed by this real case, I bet.

In short: packagers push to stable because testing/feedback is not
efficient.

Rather than just disabling the feature, I would prefer "packager
education" and perhaps "tester education" too, like pushing guidelines,
for example.

- Making testing+feedback easier will probably provide encouragments in
pushing to test rather than to stable.
- Testing could be slightly improved by publishing new test packages and
the reason for them +(BZ link if some) on some easily accessible webpage
(plus an RSS feed, maybe). Of course, this would be efficient only if
the "bandwidth" of such an info source is reasonable.
- Bumping the karma according to a reply in BZ could also probably
improve testing feedback.

> If you can think of more, please post them! But even if you just agree
with me, please reply so the other FESCo members don't think it's just
me!

I think the current situation does not satisfy testing feedback
requirement and as a conclusion, I definitely give you my voice.

Patrick