Re: [gentoo-dev] where goes Gentoo?

[Aron Griffis]

Note: I've snipped a lot of quoted text below, but took full context
into account in my replies...

Lance Albertson wrote:[Mon Jun 06 2005, 09:02:21PM EDT]
> I'd say as a global goal, yes I'd agree with you. Gentoo as a global
> entity should stay where its at, but that doesn't mean a subset of
> Gentoo could have a goal towards being enterprise.

I think that working on methods to use Gentoo in an enterprise setting
is cool.  I'm looking forward to seeing how people creatively solve
some of the problems I mentioned without disrupting Gentoo's core
development.  I did not mean to imply that *all* of those problems
need to be solved in order for Gentoo to be usable in an enterprise
setting.

> I don't really see Gentoo has a hobbyist distribution as a whole.

Sorry if it seemed like I was putting Gentoo in a box.  That wasn't my
intent.  I wasn't using the term "hobbyist" derogatorily, in case that
wasn't clear.

> I envision the 'server/enterprise' project to help create numerous
> tools that help aide Gentoo in a production environment. There's
> a lot of cool stuff we could do to help make it run better in that
> type of environment.

Totally.  In fact, some of the same tools that would help enterprise
users would also be useful to ordinary users.

> > And that is why Gentoo exists for the developers first, the users
> > second.
>
> I see your point there, but I also think theres a group of people
> that also like gentoo in the enterprise realm. I remember at the
> last LWE show in San Francisco, there were numerous people asking
> about Gentoo and making it more 'stable'. This would really be tied
> to an enterprise level of Gentoo. So I know there is interest out
> there. We all have opinions on were Gentoo should fit in, so I don't
> see why we couldn't fit there.

If there are users wanting that, then I look forward to seeing them
step up to the plate and help to solve the problems.  I'd reiterate,
though, that the solutions need to be creative enough that they don't
disrupt Gentoo's core development.

What do I mean by that?  Let me give you an example: multilib support.
Jeremy and others have been working on this for a while now.  They've
gone through a couple iterations of efforts, but take a look at
Jeremy's blog and you'll see that he acknowledges it needs some
reworking.  Despite that, the default amd64 profile is multilib, and
there is no higher indication of stability in portage than that.

How should the enterprise subproject approach this problem?  Options:
(1) They could raise a fuss that Jeremy has taken amd64 down this path
before the technology was ready.  After all, that makes a mess of
enterprise-readiness, because any consumer is going to need to migrate
to the next multilib attempt in the future.  (2) They could continue
to rely on the non-multilib profile and wait for the multilib
implementation to stabilize into something that isn't going to keep
seeing big changes.

IMHO the best approach is (2).  It leaves the default amd64 profile
multilib, which is fine for most users.  It is more work for the
enterprise subproject, but allows Jeremy to continue his development
unhindered.

Disclaimer: I don't know that much about multilib or the current state
of its development.  If I've mischaracterized it, I apologize in
advance.  My intent was to present a possible scenario and explain my
reasoning why I hope nobody will try to retarget the core of Gentoo
development at the enterprise.

> Anyways, you made some great points on where we fall, but I don't
> think we should shoot down the idea or potential because some of us
> don't think it'd work.

I agree with you.

Corey Shields wrote:[Mon Jun 06 2005, 10:18:31PM EDT]
> I don't feel that the list of requirements you have for "enterprise"
> linux is necessarily what the enterprise needs..
>
> I think Gentoo has some steps that can be taken to be a better
> enterprise player, but to come out and state that it won't work is
> a bit bold.

Ah, sorry, that isn't quite what I meant.  Rather I intended to point
out that we should not be deluded into thinking that the changes
required for Gentoo to be enterprise-ready are small.  Some of the
changes are surmountable, but each one could appear to necessitate,
IMHO, a change at the core of Gentoo development.  I would prefer for
the solutions to be possible more transparently.

For example, one way a company could presently deploy Gentoo
internally would be to (1) make a snapshot of the portage tree and
deploy based on that, (2) manually backport bug- and security-fixes to
their snapshot.  Sometimes the manual backport would be easy,
sometimes it would be more difficult, and sometimes the decision would
be made to move forward on a given package version.

In other words, a company can implement a Gentoo product lifecycle
as a wrapper around the existing Gentoo development process.  It is
a lot of work for the company, and they'd better hire some bright
sysadmins, but it would be possible.

If there is an enterprise subproject formed in Gentoo, I'd like to see
their methods be similar.  Develop tools that make it easier to manage
and maintain an enterprise deployment, without changing how Gentoo is
currently developed.  Without hoisting new expectations on the Gentoo
developers, release process, etc.

> Wow...  as a sysadmin who has run Gentoo in some very high profile
> production systems that's a bit offensive to think I used it outside
> of a hobbyist platform..  IBM didn't just donate a $30k system for
> ppc64 development to make it better for someone's basement use, so
> I don't think I'm alone in thinking that Gentoo is above "hobbyist".

I did not intend "hobbyist" to be disparaging.  I think that the big
companies (including HP, who has also donated tens of thousands of
dollars of equipment btw) see a lot of potential in Gentoo.

> Gentoo is already a fun distribution..  I don't think that has to
> change to meet enterprise goals.

Great!  I think we are closer in our perspectives than it seems.

Regards,
Aron

--
Aron Griffis
Gentoo Linux Developer