Re: [gentoo-dev] Gentoo's problems (was: Introducing the Proctors - Draft Code of Conduct for Gentoo)

[Ciaran McCreesh]

On Wed, 14 Mar 2007 19:30:37 +0100 Alexandre Buisse
<nattfodd@g.o> wrote:
> I quite agree with the Patriot act comparison, and I would be
> interested to know what you think our real problems are.

Not a complete list, but probably a good starting point:

* Portage. Gentoo hasn't delivered anything useful or cool for two
years or so. Things like layman are merely workarounds for severe
Portage limitations (not a criticism of layman). Delivery to end users
is based around what's possible with Portage, not what people want or
need. In the mean time, managing a Gentoo system has become much more
complicated due to the increased number of packages on a typical system
and the increased requirements for the average user. Combined with
serious improvements in the competition, Gentoo's benefits are rapidly
diminishing. Until there's a general admission that Portage is severely
holding Gentoo back, anything delivered by Gentoo will be far below
what could really be done.

It's been claimed that Gentoo lacks direction. It's more accurate to
say that the inability to change Portage prevents Gentoo from going
anywhere. That small interface improvements can be passed off as a big
deal and that users get excited over minor config file tweaks is
indicative of how low people's expectations really are.

I don't claim to know everything that users want from the package
manager. I know that everything in [1] has been described by at least
one user as a major advantage for not using Portage. Unfortunately,
most of these aren't things that can be delivered easily with the
current codebase.

(Incidentally, since someone will probably try this argument: I held
these beliefs long before I started work on a Portage alternative.)

* Similarly, the belief that Portage defines Gentoo and represents a
lot of work. The tree defines Gentoo, and contains far more code than a
mere package manager.

* Low QA expectations. Gentoo's QA isn't any worse than it was two
years ago. However, expectations are much higher due to improvements in
other distributions, and the increase in tree complexity makes
mistakes much more severe.

Mistakes can be classified as those that can be detected automatically
(things are improving in this area -- for one example, adjutrix is being
used to detect forced downgrades), and those that can't. Reducing the
latter involves education and ensuring that developers are aware of
expectations -- developers shouldn't be relying upon the QA team to do
QA.

Unfortunately, some developers simply won't fix QA mistakes. When
something like this happens:

11:16:24 <@genstef> hansmi: bah fix your qa stuff yourself if you think
I am wrong. I wont do something I dont agree with

something has to be done to prevent the developer in question from
continuing to hurt the users.

* The wrong idea of what the user base is, and what the target user
base is. Gentoo's direction is too heavily influenced by a small number
of extremely noisy ricer forum users, many of whom don't even run
Gentoo. Unfortunately, this self-perpetuating clique wields huge
amounts of influence.

* The repeated abuse of silly phrases like "Gentoo is about choice",
"Gentoo is about the community" and "Gentoo should be about fun" to
attempt to rationalise insane policy decisions. Choice, community and
fun are all very well, but without a quality distribution they're
worthless. The primary goal should be a good distribution, with the
rest as things that come about as a result.

* Finally, of course, the widespread refusal to accept what the real
problems are, when it's much easier to blame everything upon a few
people or groups. It might be nice and easy to think that Saddam has
weapons of mass destruction and is secretly harbouring Bin Laden,
particularly when a few disreputable news channels are going around
saying it's true, but we all know how acting upon such delusions works
out...

[1]: http://ciaranm.org/show_post/95

-- 
Ciaran McCreesh
Mail                                : ciaranm at ciaranm.org
Web                                 : http://ciaranm.org/
Paludis, the secure package manager : http://paludis.pioto.org/