[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Process is no substitute for understanding

On Sat, Mar 25, 2000 at 03:54:18PM +0000, Ian Jackson wrote:
> Anthony Towns writes ("Re: Process is no substitute for understanding"):
> > Elitism isn't exactly Right: you want the system biassed in favour of
> > good ideas (and good arguments and rebuttals), not in favour of smart
> > people. Sure, any successful system will *look* like it's biassed in
> > favour of smart people, but that's only because they come up with more
> > of the smart ideas. It's a correlation effect, not an engraven-in-stone
> > TRVTH.
> I think that by biassing the system in favour of smart people we will
> bias it in favour of good arguments and right answers.  Ie, we can
> make the correlation between smart people and right answers work for
> us by putting an expert in charge.

That's mostly true; the point of saying it's a correlation not a direct
relationship is just to say that conceivably other methods could be as
or more effective.

> > And the problems with engraving a couple of names in stone as the be-all
> > and end-all Masters of Policy is that when they get sick of it, there's
> > no easy way for someone to take it over ("Hey, you're not elite enough to
> > do the job!", "I don't have enough time to do all this", and so on),
> Not at all.  The Project Leader can - and should - replace or
> supplement policy maintainers who are not doing a good job either
> because of temporary or indefinite lack of available effort or for any
> other reason.  There's no reason that has to be difficult.

But it tends to turn out to be difficult anyway. Witness what happened
when Christian left: policy was essentially dormant for five months, and
we had to make up a whole new process to use when no one was really ready
to step forward as the new Czar. Witness what happened when James and Joey
got sick of doing new-maintainer, we spent God-knows how long arguing
about it, and saying how all the people that did volunteer weren't good
enough, and the only way we've finally started to try to get something
done is again by reworking the whole process and giving the "non-elite"
some way of helping matters go through (although, IMAO what we've ended
up with is just a more bureaucratic version of what we started with,
but hey...)

So there's no obvious reason why it *has* to be difficult, but it certainly
seems to turn out that way.

> >  and when the Czar actually makes a mistake there's no easy checks
> > and balances so s/he can just say "Ooops" and avoid masses of
> > flames.
> Of course the policy maintainers can just change their mind !  

If the Czar makes a mistake, there's a good chance it's already made
its way into official policy and some bleeding edge maintainers have
already implemented it, or it's been outright rejected and the proposers
are already sharpening their knives and complaining to the tech ctte. If
J Random Debianer makes a mistake, it just means some proposal needs a
quick rewrite, and no one really needs to know or care.

> Now
> perhaps we've got into a culture in the project where people are
> discouraged from changing their mind, but that would be a separate
> problem.

I don't think this is the case at all, happily.

> If a mistake has been made the easiest way to fix it will be just to
> persuade the policy maintainers that everyone disagrees.  If the
> policy maintainer refuses to budge in the face of unanimous opposition
> they should probably be replaced, surely ? 

Why would they need to be replaced if they're the best man for the job
for everything but this one issue?

> Of course if the
> opposition isn't unanimous then perhaps there is a real technical
> dispute at stake, and if the discussion is no longer fruitful then the
> technical committee should decide.

Or, even more likely, there's a *political* issue at stake. Consider the
`a package can only have one maintainer' debate that raged a while ago.
There isn't really a technical issue there, just a political one (in
the sense that the only differences are related to people, not to the
distribution itself).

And the correct solution there seems to have been just letting the
discussion die until it became obvious that multi-maintainer packages
aren't *that* bad (we now seem to have boot-floppies, dpkg, gcc (well,
it goes to a list, anyway :), mozilla, policy, mc, and tetex), and then
just happily change policy. Trying to take it to the tech ctte or general
resolution would probably have just continued the division.

> For any kind of decisionmaking there will always be hotheads who claim
> that someone is being an obnoxious dictator.

You keep using words like "hotheads" and "weenies" as though once someone
is a twit, they're always a twit. It's not true. Some people are hotheads
on some issues, weenies on others, and experts on the rest. Most people
are, IMO.

You said in another message that the bathwater was the often pointless
bureacracy that goes with the current process. The baby is making
decisions by concensus, as opposed to an advised dictatorship. (Clever
baby, hey?)

Maybe we should try just changing the bathwater instead of throwing
everything out.


Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG encrypted mail preferred.

 ``The thing is: trying to be too generic is EVIL. It's stupid, it 
        results in slower code, and it results in more bugs.''
                                        -- Linus Torvalds

Attachment: pgpgeY47VguaY.pgp
Description: PGP signature

Reply to: