systemd and changes

[Rudolf Kastl]

2010/8/23 drago01 <drago01 at gmail.com>:
> On Mon, Aug 23, 2010 at 11:22 PM, Mike McGrath <mmcgrath at redhat.com> wrote:
>> On Mon, 23 Aug 2010, Bill Nottingham wrote:
>>
>>> Mike McGrath (mmcgrath at redhat.com) said:
>>> > > My concern with this line of thinking is that you're asking us to quantify
>>> > > the unknown unknown, and define a time period of testing which is
>>> > > 'long enough' for us to catch all the unknown unknowns. This seems
>>> > > impractical, in as much as it doesn't give us any clear criteria to define
>>> > > success with.
>>> >
>>> > It's just risk management.  I think we'd be better off acknowledging there
>>> > are unknown unknowns and try to mitigate them.
>>>
>>> Sure, but when you say 'we should hold off X period of time' in order to
>>> mitigate unknown unknowns, how do you define 'X'? How do you know when it's
>>> ready? All I'm seeing are appeals to gut feelings. We can all say that 'more
>>> time == more testing', but how do you claim 'good enough'?
>>>
>>
>> I'd say one release is good enough for Fedora.
>>
>>> > ready.  Unfortunately that's not the path we seem to be on.  We unwisely
>>> > seemed to declare it ready before anyone even saw it then we ignored what
>>> > we didn't know as if we knew there were going to be no problems.  The sad
>>> > thing is that's such an easy fix by making brand new features for core
>>> > components like this opt in, even if it's just for a single release.
>>>
>>> Having to support multiple boot paths for the system, making everyone
>>> who gets odd bugs filed against kernel, dracut, plymouth, etc. triage them
>>> isn't exactly an 'easy fix' - it *adds* complication to both paths.
>>>
>>
>> I'd rather have multiple boot paths to choose from then only one boot path
>> that is 2 months old.
>
> Being "2 months old" isn't a problem in itself ... bugs on the other
> hand might be if they can't be fixed in time (this does not include
> already fixed ones).

it is, because 2 months for a critical system that is required for
booting doesent provide enough time for testing at all. just note that
after reinstalling systemd on rawhide it now boots again (the
introduced bugs with linking havent been properly fixed by package
updates but only with a reinstall since the workarounds published on
the list didnt work for me at all) , but for some reason
networkmanager didnt start automatically at boot. a problem found in a
few seconds of looking at it. a change like that needs a longer
testing period. i agree that fieldtesting is a really good way of
finding alot bugs quickly and getting them fixed but that is really
what rawhide is there for. Sorry for not beeing that optimistic after
working with software for a rather long time.

kind regards,
Rudolf Kastl




> --
> devel mailing list
> devel at lists.fedoraproject.org
> https://admin.fedoraproject.org/mailman/listinfo/devel
>