Re: [gentoo-dev] Gentoo part II.

[Paul de Vrieze]

On Tuesday 15 July 2003 12:37, splite wrote:
> On Tue, Jul 15, 2003 at 11:06:09AM +0200, Paul de Vrieze wrote:
> Content-Description: signed data
>

> > developers for gentoo. Having a single "boss", and a "lieutenant" with no
> > structure at all is not going to work. Especially as the amount of
> > developers grows.
>
> Works for the Linux kernel.  Why do you need more developers?  Does every
> package in the universe have to end up in the portage tree, with its own
> developer?  I'm quite serious.  Just because someone cobbles up an ebuild
> for whatever obscure package, does it have to go in?

Well, Linus has more then one luitennant. And those luitennants also have 
their own sergeants etc. The kernel development process is surely structured. 
It has local responsibilities just as gentoo is going to have.

>
> > We need structure.
>
> You have structure now.  Why make it a full-blown bureaucracy?

No, we don't (well didn't, as we are trying to set it up now)


>
> > Part of that structure is a place where things are documented, like
> > responsibilities.
>
> You already have a place for documentation.

We need developer-developer documentation, and indeed we have a place for it. 
Having a place though, does not make that it doesn't need to get written.

>
> > The problem is that with 30 developers you could easilly ask something
> > you didn't know. Now the problem is that you can still ask, but you don't
> > know who to ask. Documenting procedure and formalizing a bit should help.
>
> What's wrong with giving a shout-out to gentoo-dev?  I'd rather see someone
> documenting Portage better (say, "how SLOTs work"), than documenting
> procedures for asking questions.
>

We are not documenting procedures at all, or aiming to. We are documenting who 
does what, so we can see the gaps and overlaps.

> > There are also many people and organizations that want gentoo to run on
> > their servers. Those people have one thing they REALLY REALLY hate, and
> > that is comming to office in the morning and finding out that the nightly
> > world update fucked up their setup, and it will take at least until the
> > end of the
>
> Then those people shouldn't be idiots.  Seriously, who in their right mind
> runs automated nightly updates on production systems?  Run them on a test
> machine, then if things look okay afterward, push it out to the production
> boxes.  That's common sense.
>

Of course they don't do that, but there are contacts with corporations that do 
want to use gentoo because they like its structure, but don't like the moving 
target nature of the tree. The idea is to create releases to suit those 
users. Those releases then only receive security fixes and major bug fixes.

> > morning fixing things up. Normally such thing will mean a great loss of
> > productivity.
>
> Then they should be paying Red Hat or SuSE for support.  Folks, you're
> not responsible for someone being dumb enough to let their systems be
> borked nightly.  In fact, if you start acting like you are, you may
> well find someone trying to hold you to it in court.
>

Of course no smart person does a nightly update on a production system. They 
just want MORE stability.

> > Since we believe that the gentoo technology is better than the
> > competition,
>
> Gentoo tech is quite nice, but it doesn't have to be all things to all
> people.
>
I didn't say it is perfect, I said that it is better than competition for the 
area's we care about.

> > even for servers we want to offer what they want while keeping what we
> > have.
>
> If they want a system that's flexible and easy to fix and customize,
> Gentoo's great.  If they want guaranteed uptime, they should buy a
> commercial distro and a service contract.
>

No OS vendor guarantees uptime as there are too many untracable causes for 
crashes. Many of them hardware (or position of the moon) related.

> > For offering what is needed for servers we do need more quality
> > assurance.
>
> I hate to keep bringing up Debian, but it's a perfect example.  Their
> desire for QA has brought the project to a virtual standstill.  Even given
> the huge number of Debian Developers, they can't validate all 10,000+
> packages on the 11 architectures they support in any timely fashion.
> That's why they're taking years between releases now.
>
> As long as Gentoo keeps being "good enough", it will have users.  As long
> as its developers take pride and derive enjoyment from working on it,
> Gentoo will be good enough.  You don't need Quality Assurance committees
> drawing up charts and setting milestones.  If you want to set a release
> date, just pick one, Bach's birthday, whatever.  Ship whatever you have
> on that date; it's still bound to be better than Debian or Red Hat, even
> with all their QA aparatus.

There is no final decision on QA yet, but the idea is to pick a date, say 
Bach's birthday, then do a fixed period of testing and fixing (coming from a 
stable branch that should have no bugs anyway), then release.

>
> > With QA and the growth of the project comes a management structure. That
>
> Any "project" has a management structure, by definition.  If the present
> structure can't keep up with growth, another possibility is to check the
> growth.

No growth implies a change of the structure. We are currently in that process. 
And indeed we do check growth rates with our ability to manage it.

>
> > structure is inevitable. John made a proposal on how to arange parts of
> > that structure. While we will put every effort in it not to create a new
> > debian, we need to be more organized than before.
>
> That's appreciated, but I and a few others think he went over the top.
> Debian is not the model to emulate.  I wouldn't even try to make a "fixed"
> version.  Maybe Gentoo should stick to its founding spirit and come up
> with something different.
>
I never said I agree with everything that John said. 


> > So please all discuss the merrits of his proposals. I believe that the
> > problems they try to address are there and are well accepted.
>
> I don't think the actual problems have really been discussed, at least
> not here.  As I asked before, whose needs aren't being met here?  Simply
> stating "we need structure" like it's axiomatic isn't an argument.  If the
> developers are having fun, and the users are getting something useful (and
> for free), why isn't that sufficient?
>

Because we want to be better tomorrow than we are today. Gentoo is not 
perfect, and probably newer will be, but we can aim towards it. For gentoo 
developing to stay fun, developers must keep the feeling that their ideas are 
being considered. Any good idea that stays in a developers/users private 
overlay is basically a waste. Today that happens too often because of certain 
chokepoints.

Paul

-- 
Paul de Vrieze
Researcher
Mail: pauldv@××××××.nl
Homepage: http://www.devrieze.net