Anders Karlsson wrote:
* Andrew Farris <lordmorgul@xxxxxxxxx> [20080310 13:18]:
Anders Karlsson wrote:
* Andrew Farris <lordmorgul@xxxxxxxxx> [20080310 12:18]:
[snip]
Again.. as has been stated previously by people with authority around
here (which I am not)... testers did not provide proper feedback on this
kernel, via the tooling in place for it [1]. Proper QA is not the
responsibility of someone else. It is the responsibility of the
community. People's lack of participation in how the system works are
their own failure in this case.
So the Fedora process is;
"If *we* break it, it's *your* fault because *you* did not test it" ?
I stand by my earlier comment that this shows little but contempt for
the userbase.
Fedora devs have *countless* times proven they care what the users think
about this type of situation, but users who only share what they think
after shit breaks are useless to everyone.
Oh, I'm sorry. I thought that the idea was to have a three-tier system
with stable, testing and unstable repos (to borrow terms from Debian),
where the idea is to run stable if you intend on doing something
productive, testing if you participate in QA and unstable if you like
living on the edge.
That is in inaccurate picture which probably has alot to do with your
frustration over this. Debian's unstable and testing are completely
separate repos not intended to get flowed together at a regular timeframe.
Fedoras are meant to be very temporary, for testing. Updates-testing is a
place where things desperately need testing interaction so they can be
pushed because they are only there because they fix known bugs... not
because its new stuff that happens to be unstable and is not getting
dropped into the primary repo.
As a suggestion to improve things then, why not regularly have a
roll-call on the standard fedora-list asking for volunteers to
participate in using the updates-testing repo? Preferably with a short
few lines about "how to enable it, and how to disable it again should
you wish to stop participating" ?
Good ideas.
Pushing known broken packages from testing to stable is not
acceptable, just because it did not receive enough QA in testing. Yes,
you get what you pay for, but work ethics should prevent these things
from happening.
It was known to fix some bugs. Feedback was not provided about having
major breakage. What exactly is the proper procedure in your eyes here?
Not push the kernel which is known to fix bugs? I'm really curious.
Fixing bugs is good, but pushing broken packages is bad. I'm not
saying bugs should not be fixed, but knowingly introducing breaks
elsewhere seems to be putting the cart before the horse a bit. If the
bugs fixed were security related, or critical (known data corruptor or
similar), I can see the need.
I agree, 100%. Broken packages shouldn't be shipped. The question is whether
something is seriously broken, or happens to be broken for 1 or 2 people with
strange hardware, odd configurations, or transient problems. Keeping in mind
this is not a distro intended to fit into the 'nothing will ever break'
category. The front page of fedoraproject.org should make that one pretty
clear. The minimal testing feedback received in that case would have made me (a
tester) think it was perfectly reasonable to push it... up to the time the push
comment was posted to bodhi.
Proper process... Well, I'm just an end-user, so my opinion does not
weigh much. I would however expect that the process in place to handle
the situation, and in all fairness - it does seem to work quite well
most of the time.
Fedora 8 is released, it is supposed to be the "stable" branch, or are
you trying to tell me that Fedora 8 is still in development, even with
Fedora 9 pending? If so, are users supposed to run Fedora 7 to ensure
they are not being treated as lab-rats?
You can answer this question, see above. Do you want known bugs to get
fixed? How many weeks do you want to wait for them? Would you prefer
nothing get released until its had an 'adequate' testing, and never get
released if not? Do you realize that because there are obviously not
enough testers this means that *most* updates will never get released?
Do I want bugs fixed - yes, but not with the pricetag of breaking
other functionality *knowingly*.
What you are describing is a problem with packages not getting
adequate exposure to testing - and I have made a suggestion earlier in
this mail about what could be done to help that.
I agree, more testing exposure *AND* participation is needed and is the real
root cause of this problem. Updates cannot be shipped knowing they are valid
unless someone tests them, and thats the point of my comments up to now. The
primary assumption of using Fedora is that it is as good as the community helps
make it... a drastically different scenario to RHEL and its important to understand.
Welcome to CentOS (just go install it now). That is not what Fedora is
supposed to be about, at least from my perspective and why I spend my time
with it.
I've got CentOS equivalent running on some systems, and for my X60s I
went with F8 for driver reasons. I'm just surprised about the shift in
focus in Fedora, that's all.
The kernel update did not eat any data for me, it just broke X for
me. Rolling back kernel, problem solved.
I must be incredibly dense - so I'd appreciate a thorough explanation
about this situation just so I know what to do to avoid having my
systems sabotagued in future. If I understand you right - you are
saying that Fedora does not want "just users" because they are totally
useless and just a burden to the project?
That is not at all what I said, or meant. What I said is people with
nothing but complaints after the fact are helping nothing. That does not
mean they cannot use the distro, nor does it mean the community is not
*intended* to work for them and prevent this. What it means is they are
not part of solution or prevention. Thats it. The community as a whole
should have prevented this; yet in this case it failed to do so. A
situation to be learned from, not one to pack up camp and run away from.
And what else are the "just users" supposed to do, after the event,
when they sit with a system that for some reason no longer will let
them start X, use their wireless network or suspend/resume? They *are*
reporting the problem because someone broke something for them,
something that used to work. That's called 'regression', and that is
usually taken quite seriously in software development.
I think its bad, very bad. But the fact that it happened is not the end of the
world either. The regression can be fixed (unregressed!). It happens in
commercial software at *regular intervals* as well (although, the linux
community should be better than that).
If the required steps are to get the packages exposed to more testers,
then I'd expect steps to be taken to recruit more testers. Advertise
the updates-testing repo more, encourage its use, prod the mailing
lists regularly for participants. Yes, we should learn from this, as
that is what good organisations do. Leverage the information in Smolt
as well, maybe have a push with that, allowing people to tick a box
saying that "Yeah, I could be interested in doing some QA for you"
when submitting the hardware profile and then use that to get as much
coverage of hardware as possible.
Thanks!
/Anders
That is excellent feedback and I hope there are documentation people taking
notice of this thread.
--
Andrew Farris <lordmorgul@xxxxxxxxx> www.lordmorgul.net
gpg 0xC99B1DF3 fingerprint CDEC 6FAD BA27 40DF 707E A2E0 F0F6 E622 C99B 1DF3
No one now has, and no one will ever again get, the big picture. - Daniel Geer
---- ----
--
fedora-devel-list mailing list
fedora-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-devel-list
