[CentOS-devel] progress?

[Fabian Arrotin]

On 02/20/2011 07:06 PM, Dag Wieers wrote:
> On Sun, 20 Feb 2011, Fabian Arrotin wrote:
>
>> On 02/20/2011 06:13 PM, Dag Wieers wrote:
>>>
>>> So it's absolutely normal a CentOS 5.6 release is 10 weeks late and
>>> there's no intent to speed up the process to lower the time to do future
>>> releases ?
>>>
>>> If that is the case, we might want to make that more clear on the wiki,
>>> eg. in my CentOS introduction presentation from 2008 we still mention that
>>> releases are (up to) 4 weeks behind RHEL.
>>>
>>> I don't think most of the users ever expected to be without security
>>> updates for 10 weeks or more when choosing CentOS, and that is an
>>> important characteristic.
>>
>> Right, but as you said so in your multiple CentOS presentations, there
>> are no warranty nor SLA on a delivery time : IIRC you explained that
>> very clearly in the Pros/Cons slides between RHEL and CentOS ;-)
>
> Fabian,
>
> I did say that was a con for CentOS, but we did mention that security
> updates may be delayed for 48 hours and releases may be delayed up to 4
> weeks. Both are no longer the case. We have had security updates delayed
> for 3 months (!) and a whole release delayed for 3 months since I wrote
> that presentation.

*nod*

>
> I haven't done any CentOS presentation since I left the team, but it is
> undeniable that despite any automation work since then releases only take
> longer.
>
>
>> Working a little bit behind the scene and knowing the CentOS developers,
>> i'm quite astonished that all the people complaining about the 'delay'
>> always think about technical issues that CentOS developers have to deal
>> with and not with their personal problems they have to deal with ...
>
> True, but there are ways to overcome those problems if, and only if, we
> consider this important. Up to now this was _never_ considered important.
> Remember: "we release it when it is ready".
>
> That is not what users like to hear in this context, because it basically
> means there is no way to speed it up. There's not even a plan to look
> exactly what is the bottleneck and where can we improve it.
>
> I remember one CentOS 5 release was delayed for one additional month
> because a single developer was getting married. We all sympathized, and
> something like this can (should!) happen ;-) But shouldn't the project
> foresee and/or protect the process from such occasions ?

Also true indeed ...

>
> In 2007 CentOS had a lower limit of about 5 million installations, that
> might be a lower limit of 10 million today. Maybe even a multitude ?
>
>
>> And on each release it's the same thread coming. It seems to me like a
>> 'chicken and eggs' problem : what has to be solved first ? people
>> wanting a faster distro release or documenting it first ? Don't
>> misunderstand me : i really would like to see the whole process
>> documented and i think that some other members agree on that fact. (It
>> seems SL team documents more what they are doing for example). But here
>> is now a simple fact : consider how much time Johnny took to answer all
>> the same questions (and even answered multiple times with the same
>> answer) and so the time he couldn't spend on the build process itself ...
>
> That argument has been used in the past too to silence any critical voice.
> But as soon as the release is out, nothing seems to really happen despite
> any promises. And the delay is becoming longer. We used to be able to
> release CentOS-5 release in one month, remember ?

Yes and once again, don't get me wrong : i'm not arguing against any 
critical voice or to silence comments. The fact is that $now doesn't 
seem (once again) the right time to do it. I know you'll answer directly 
to that with a 'so when is a right time to discuss that again ?' and 
i'll answer (still *my* opinion and not the project's answer) that just 
after the release would be a good time. This is a very particular time 
and i think that it's the first time in the project lifetime that there 
are 3 releases to cover at the same time (4.9, 5.6 and 6.0)

>
> There's no reason that CentOS 5.6 is any more difficult than CentOS 5.0,
> while CentOS 5.0 (a major release !) only took 1 month to produce. Sure
> CentOS 6.0 is a lot bigger, but CentOS 5.6 is not, and CentOS 4.8
> certainly was not.
>

Yeah and one thing we can all agree on is the fact that this time for 
6.0 we tried to open the process and let people participate and that it 
was a big fail ....

My own conclusion (still *my* opinion and not the project's position on 
that question) is that most people just want to have a 'cut-and-paste' 
solution to automate their own rebuild, instead of seeing CentOS to 
release a distro faster.
I know i've already said that in that thread (but a lot of people have 
'echoed' their own answers too, right ? :-) ) but i've had personally 
the case where people were asking to 'help the project' and when they 
were pointed to either improving the website, translate the wiki, 
chasing after potential banding issues, etc, etc .. the only answer i've 
got *multiple* times was "no, i'm not interested in doing that : i just 
want to rebuild packages" .. so each time it proved me that such people 
aren't interested in helping the project as a whole, but instead just 
want to focus on build issues. I'm really wondering (and still *my* 
opinion) if those people are interested in CentOS as a project, or just 
want to 'suck' some build scripts (which are just wrappers around 
mock/plague as stated so much times in that thread) to produce their own 
respins.
Once again, don't get me wrong : documenting that would be very 
fantastic (still my opinion) and that's what i would like to see from a 
FLOSS project (but Red Hat doesn't do it either, otherwise it would be 
'too easy' to rebuild it :-) )

Fabian