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

Re: Candidate questions/musings

On Mon, Mar 22, 2004 at 06:43:00PM +0100, David N. Welton wrote:
> Anthony Towns <aj@azure.humbug.org.au> writes:
> > > Frankly, the most exciting development in Debian I've seen lately
> > > is Bruce Perens' UserLinux, [...]
> > For example, as well as the many things Bruce came up with that
> > worked out, he also proposed things like switching us to rpm, and
> > dropping packages like "bitchx" to make the distribution more
> > approachable to more people, or the Debian chicken, or the "deity"
> > codename, or whatever else. Bruce was, effectively, hounded out of
> > the project two or three times for his efforts.
> He eventually left in something of a hissy fit, but aside from that,
> yes people disagreed with him, and there were flamewars, but he did
> get a lot of good stuff done, and even where one might disagree with
> him, at least he took a stand for something he felt would improve the
> project.  That's different from "coordinating".

Yes, it is. And Debian's response to that was unpleasant enough that Bruce
decided to leave two or three times. That other people don't want to put
up with that degree of unpleasantness and choose to "coordinate" instead of
"lead" shouldn't be particularly surprising.

> > > [...] Not that I'm not grateful [...], just... I want a coherent core +
> > > aptable addons.
> > So the problem with this is still the same as it's always been:
> > "aptable addons" and the idea that you can split Debian into
> > distinct components and work on the independently doesn't really
> > work:
> Ok, I'll take your word for it, let's not go off on a technical
> discussion - that was an aside in any case.

Getting sidetracked by a technical discussion would be such a pity,
wouldn't it?

In any event, there are other directions we can go here; the one in which
I'd like to travel is to have testing be treated as a real distribution:
one that's installable, and that's fully supported. Unfortunately that's
also not a direction in which Debian's really been willing to be lead,
and since it's not something I can do alone, it hasn't happened. People
are plenty willing to come up with their own ideas like you've just done,
or to express what they'd like to have happen, but actually following
through on the not-so-fun bits doesn't tend to happen so much.

Seriously: you say you want to be lead; but are you willing to follow? If
you want to be able to reliably run a current Debian system, it's not
particularly difficult to make testing fulfill that need: it doesn't need
much more than security support. And fulfilling that need -- particularly
if testing can be made installable, rather than merely upgradable --
removes to unpredictability of our release cycle trivially. But while
the infrastructure is all there to make it fairly straightforward, it's
still a hard job, and none of the offers we've had so far have panned
out at all.

> > When the people who do try to actively lead the project, or defend
> > its principles in a visible way get beaten into the ground at every
> > opportunity, [...]
> Can you cite some examples of people post-Bruce getting beaten into
> the ground?  My memory is a bit fuzzy, but I recall:
> Ian J - constitution and not being very present

Constitution. Created the tech ctte. Showed us how to cope without Bruce
as DPL.

> Wickhert A - nice guy, glory days of .com.  Accomplished?

New logo, and logo license.
Revived dpkg maintenance.
Brought up the "split non-free to a separate server" issue.
Debconf spec.
Ressurrected the new maintainer process.

> Ben C - don't remember much at all

Crypto in main (although HP and ftpmaster and others get most of the credit)
Appointed Manoj as secretary.
Appointed Wichert Akkerman and Jason Gunthorpe to tech ctte.
AFAIK did a bunch of stuff working with sponsors, but I try to avoid knowing
anything about that sort of thing.

> Bdale - nice, competent guy who represented Debian at all kinds of
>         conferences.  Accomplished?

Strongly supported Debian as a multiarch distro; with mips, mipsel,
hppa, ia64 and s390 getting added to woody.
Strongly supported the LSB, which missed 3.0r0, and got mishandled as far as
getting included in any of the minor updates. Has made it for sarge though.
Inspired a bunch of people to work on desktoppy stuff.
Worked with sponsors to get regular debconfs properly funded.

> Martin - nice, competent guy who is doing a good job representing
>          Debian, and, from what I've seen, "coordinating".

You can put it in scare quotes all you like, but coordination is
pretty much exactly what he proposed to do. We've had a "debcamp"
during his term and a "d-i camp", SPI's finally become able to accept
electronic donations, debian-qa's active again, pretty much under Martin's
leadership, new-maintainer summaries are sporadic instead of non-existant,
and he's politely prodded at least a few teams to make sure they do a
better job.

It's often said that people get involved in open source development
because of the reputation they can build up -- doing stuff as DPL seems
to result in nothing better than "nice, competent guy". Motivating, no?

> Somewhere along the way, apt happened thanks originally to Jason
> Gunthorpe, pools happened, the BTS evolved, the infrastructure
> evolved...  What else?

apt (ne deity) started in Bruce's time.

Pools were Bdale's idea, with detailed design by me, Jason Gunthorpe,
and James Troup, and implementation mostly by James. Guy Maor did some
design related stuff, but I can't remember if it ended up mattering
at all. Drake Diedrich made a competing implementation which wasn't
backwards compatible with dinstall, and got mostly ignored, inspiring
some bitterness on Drake's behalf; although I think some ideas got
carried over.

The BTS was written by Ian Jackson (in a language affectionately known
as iwj-perl), taken over by Darren Benham at some point who did some
packaging work on other maintenance, Josip and Adam joined at some point,
I wrote some CGI scripts as a proof of concept which then got mainlined
when the static HTML generation started killing master and setup the
archiving and started backing up closed reports, and we eventually invited
Colin Watson to join when he started writing some interesting patches; he's
now changed the .status format to something that's extensible and useful.

We have lots of infrastructure; I'm not going to even try giving a
complete rundown of how all of it was developed over the past couple
of years.

Apart from when the people above are just forgotten, or at best remembered
as "oh yeah, nice guy, whatever", most of the above are usually complained
at -- Darren's the guy who needed to be "pressured" to return all the SPI
materials he misplaced as treasurer; James is the all purpose nefarious
Cabal member; I'm the guy who quashes dissent and abuses his power. The
primary benefit of having leaders in Debian teams is that it lets the
rest of the team avoid getting cursed at.

But you really don't need to look any further than the "non-free" debate
to find examples of people standing up for what they think's right,
and trying to lead the project in that direction, and being shot down
for being a bigot, a hypocrite, at odds with Debian's purpose or whatever.

> > Basically, as it stands people seem quite happy to say "we want more
> > leadership" or "we want more chicks in debian" or "we want more
> > transparency", or any number of other things, but by our actions at
> > least, it seems more like what we really want is somewhere we can
> > say whatever we think, and abuse anyone who doesn't think the same.
> Hrm.  I think there's some truth to that.  Debian has a lower average
> maturity level than other projects, because it's easier to get
> involved.  

If it were mostly newbies at fault, that'd be one thing, but it really
isn't. You've been around longer than I have, but can't seem to think
of anything nice to say about anything the last few DPLs have done, eg.
Craig Sanders has been around longer than both of us, I think, and has
been cited as being pretty immature quite recently.

> In any event, I
> still think a good leader could help foster the right sort of
> atmosphere (Bruce wasn't perfect in that regard).

How, exactly? One thing we haven't tried recently is a DPL who
flames people who don't do what he wants, and dismisses people who don't
agree with his policies. That doesn't strike me as "fostering the right
sort of atmosphere", but what's the alternative? Having DPLs who work with
people who disagree with them, and try to persuade them politely without
making a fuss? But then, that's back to "DPL's not doing anything".

Personally, I don't think it's a problem that the DPL should be solving,
I think it's a problem the developers need to resolve ourselves.

The "Bruce" lesson is a pretty easy one IMO: if we want to encourage
people to take leadership roles and actively develop and push agendas to
improve Debian then we need to make sure we don't have an environment
that makes leaders like that throw hissy fits and quit the project.
The reason we didn't mind Bruce's hissy fit at the time, and went with
Ian Jackson is that people who push their agenda are a real nuisance
when they choose a bad one.

AFAIK, I'm yet to see anyone willing to give up anything much in order
to change that though. See the "chicks in Debian" thread we just had on
this list, eg, which was pretty much used by every single participant
to advance their own agenda, rather than trying to work out a single
direction we can all choose.


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

             Linux.conf.au 2004 -- Because we could.
           http://conf.linux.org.au/ -- Jan 12-17, 2004

Attachment: signature.asc
Description: Digital signature

Reply to: