976_Debian_Cosmology.ogv

Okay, and now for the last talk in the morning session,

Joey Hess will talk about Debian Cosmology

[applause]

Well, thanks, good morning everybody, I hope you had a good night's sleep.

I enjoyed sleeping out in the tent, in the middle of Switzerland

looking out over the lake

this is kinda the first Debconf where I've kinda had a problem

If I just look over there I'll probably just lose focus for a bit, it's so gorgeous.

I thought this would be a good place to get up on a mountaintop, as it were

and think about the bigger picture

and try to think about some of the big questions, the big vague things we wonder about

but don't really, sometimes, talk about

maybe in public in front of a live streaming audience, I don't know.

I have this crazy Debian Cosmology idea.

And, let's look at Debian, let's look at the Universal operating system

and think about thinking back 20 years back

to when Debian was founded, up to the present

and where it's going to go from here.

Back in the beginning, there was kinda this void.

and there was a gap [applause]

and Ian Murdock saw this, and he said, well...

let's make a new Linux distribution to replace SLS

It'll be great; I'll get it done in a couple of weeks.

[laughter]

And this was back in 1993.

And just as with the big bang

you have the laws of nature somehow forming out of the void

we developed these standard principles of Debian

that have pretty much stood the test of time

although some of them like the one package, one maintainer thing have changed over time.

But this is all the stuff that we think of as the core principles of Debian today, probably.

And this was in the period '94 to '98

this early period where there weren't very many people involved in Debian

and things got done fairly quickly.

I have down here one of the initial threads for the Debian Constitution.

This is where Ian Jackson said

I think we'll use this constitution proposal to bootstrap the constitution

so we'll vote on the constitution using the principles of the constitution.

That could be a kind of controversial things to say, actually

because it's a bootstrapping problem.

But the thread actually wasn't that long

by today's standards [laughter]

for something that important.

So, this was, as I said, the early period

and then in the late 90s and early 2000s

we went through this inflation period, just like the universe blew up

got bigger and bigger

we have the nice up-and-to-the-right graph

which is the number of maintainers over time

and I don't think that this data is very good

but I'm kinda happy to see that it's started going up again in the most recent election

although that's probably also just because Zack wasn't running [laughter]

So during this inflation period we had things happen

like adding ports to Debian.

One port in '98, two ports in '99

two ports in 2000; that's two ports a year.

It's a crazy rate of change.

And then we had... all these derivatives started popping up.

We'd had Debian for Hams for a while

but we got these derivatives that you don't think of much any more

like Corel Linux, Stormix, Progeny

These are names we haven't mentioned in a while

but they were the early corporate entities saying, well,

we're going to try and do something here with Debian

and modify it

and of course many more came from there.

And another big event in this period was that apt started out.

This is one of the early threads about apt

This is about a year after it started being developed.

Everybody started trying it

and realized: oh, it actually doesn't work on my system

because I have these packages that are half-configured

I have a few broken dependencies because I just forced something at some point

And everybody tried apt, and they're like, gosh, it says my system's inconsistent

and it doesn't have apt-get -f yet

so it doesn't work.

So I thought this was an amusing thread.

It's also not really too long a thread

but here's an introductory representative message

I don't know if you can read it back there

but it's just what I said, apt-get dosn't seem to work

it says my system lacks integrity

and then Jason Gunthorpe, who wrote apt, said

I don't think I've actually seen a Debian system that has a perfect dependency setup

so that apt can actually work on it.

If you think about introducing some big new change like apt

and it doesn't work at all

and this was in April of 1998

If we then move forward one month to May of 1998.

here's somebody saying

"This makes makes me wonder if we should think about dropping this autoup script..."

"...that we're using for upgrades" (some kind of a shell script or something)

"...and switch to apt."

"autoup seems to work and maybe we shouldn't postpone Debian 2.0 for apt..."

"but autoup's a hack, and apt lets you do an entire bo to hamm upgrade in dselect."

Wow. I was kinda of surprised to see this: it turns out that I wrote that.

I had no idea that I proposed converting Debian to apt for 2.0.

It didn't actually happen in May of 1998

we had to wait a whole year until March of '99 when 2.1 came out

and this is a quote from debian-history

about apt, which I thought was a great quote:

"It established a new paradigm for package acquisition and installation"

and it really did.

If you look now at things that are basically command-line compatible with apt

or more or less command line compatible

maybe they didn't quite understand the difference between upgrade and update

There's so many of them! It's crazy.

And one of the interesting things about this list

is that if you look and see which ones of these actually do it securely

it's a really small subset.

Maybe some of them use HTTPS in some way

and have a little bit of security there, I don't know.

I didn't check them all in detail.

Of course back then apt didn't have any security either.

It was just pulling stuff via HTTP off the web

and hey, it'd be the right thing, because why wouldn't it be?

So soon after apt came out... this is a screenshot from 2002

but it was around earlier

We got apt-get.org

which was all these third-party apt repositories.

And this was kinda interesting, there were hundreds of different repositories.

You could go off, edit your sources.list, get your packages

and we kinda started thinking, wow maybe we're gonna change how Debian works in some way.

Maybe we'll have some kind of a central core

and everything else will just be pulling from other repositories somewhere.

And we kinda went off on a divergent path.

We kinda went down a wormhole to some distributed apt, or app store model

where there's Debian and all this stuff you pull in from here and there

and if somebody wants to make a package they do

and this kind of is what happened today too.

You can pull, you know, signed packages from Google

and from debian-multimedia, deb-multimedia, that kind of thing

But we didn't really go down that path.

We're still very much a centralized distribution.

I kinda think it's interesting to think about what could have happened if we'd branched off in a different way there

But there were good reasons to keep it centralized, such as security.

And if you now fast-forward to the present

here's apt-get.org from 2011

it's been broken, we can't check if these repositories work any more

we're not accepting new submissions

and this is what happened to debian-multimedia.org, which is a pity.

It's a Russian domain about motorcycles or something, I don't know.

So that's kinda the inflation period of Debian.

And then we can move forward again into the modern era.

This might be where my cosmology analogy gets a little bit strained

but we'll see.

I've picked out two things about the modern era of Debian

this past 10 years, or 15 years.

So, one of them: just as in the Universe, you have large scale structures forming,

galaxies, and larger structures.

In Debian we've kind of developed all kinds of structures

on top of the "one maintainer, one package" model

and extending it, and going beyond it.

So a few of these, such as teams...

Lucas showed us the graph of team maintenance increasing over the past ten years or so.

We've just developed all these structures

Custom Debian distributions

stuff like d-i, different projects within Debian.

So it gets pretty complicated.

It's not a heterogeneous thing - a homogenous thing

It's all clumped around in different places.

If you also look at where people are using Debian

that's differentiated a lot too.

It's not just... we are the Universal operating system, we say

but a lot of people are using Debian on servers

and a few are on laptops

and basically nobody is on a mobile phone

except for a few people who are lucky enough to still have an Openmoko, or something like that.

So, we've really differentiated Debian a lot.

So that's the large scale structure thing.

I think it's interesting to think about it

because it kinda makes you think about how Debian's evolving.

Now this is where it really gets strained.

Red shift. Okay.

[laughter]

How do we have red shift in Debian?

I don't see any red when I look out

unless I've stepped into the middle of a flame war, or something.

Here's kind of an amusing paper

which I don't think has been peer-reviewed yet.

It says, what if the universe, rather than actually expanding right now

like we think it is because of red shift

what if the mass of everything is increasing at once?

And it says, well, everything would work pretty much like it does now

we wouldn't even be able to test this theory.

And while I don't know if the mass of the universe is increasing

exponentially over time, like this paper says it is

it seems a little unlikely.

Debian's mass has definitely increased.

We have an enormous mass, and an enormous momentum.

We're moving in a certain direction

and it's really hard to move Debian into a different direction now.

So, one really easy example of this

systemd.

Think of how many threads we've had about systemd lately

and, yeah.

And this isn't replacing dpkg with apt

and breaking all of our dependencies, and having to change everything.

This is changing how systems boot

which you do once a week, or once a month, or once a year.

It's a minor change as things go, right?

And yet it's an enormous controversy inside the project.

So I think we have to think about this momentum, this mass

how do we manage it, how can we make Debian nimble

on top of all this momentum.

So I think that's probably the largest problem that Debian is facing right now

and will face in the next however far out you want to look.

It's kinda hard to give a talk about Debian cosmology

because what is a long time scale in Debian?

We have twenty years of history to look back on.

Can people think in their head, wow, will Debian be around in twenty years?

I don't know.

Pick a timescale that seems to make sense to you for the rest of this talk.

I'm not going to try and force some kind of a timescale on you.

If you want to think a hundred years ahead, great.

If you want to think ten years ahead, okay.

But I'm going to try to think about moving forward

but first I have a little digression, which I forgot about.

So, one of the examples of a way that the momentum in Debian can be a problem.

I mentioned apt.

Well there's this interesting thing being developed right now

called functional package management.

It started out with nixos

and now the GNU project has gotten involved with its...

Guix? I don't know how to say it.

The idea is that it somehow takes ideas from functional programming

and applies them to package management

so it's bread and butter for me.

I'm really interested in it being a Haskell guy now.

Being in a functional program

you're like, wow, there's some interesting ways to use these ideas.

It's not really functional, but it's a neat terminology to hook on it.

And what this lets you do, it's kind of a source based system, in a way

I don't know. Has anybody used any of these systems in the audience?

I'm just curious. You have, Zack?

I'd love to chat with you about it and get a broader idea.

The idea is kind of that you never make a destructive change to the system.

Every package change is atomic.

and if you have dependencies

you might have multiple versions of a package installed at a time.

and it's completely different than the dpkg model in every way.

And it's kind of inconceivable to think that Debian would switch to something like this model now.

It would just be so incredibly hard.

You know, switching to apt would be just nothing in comparison

and it's much later in our evolution

we have a lot more structure built up around our current system

than we did back then, even.

This is an example of something that...

The universe is coming up with neat new things

How do we possibly put them into Debian?

We can obviously package up these package managers

and make it easy enough for people to use them as a third party thing

You can install stuff in your home directory with functional package management

and just have a system on top of Debian, and that kind of thing.

But how do you integrate this kind of thing

or ideas from this kind of thing into Debian?

I think the closest we're coming is the switch to more declarative systems for Debian packages

so that rather than maintainer scripts, we have triggers, and stuff like that.

But this is just taking it to a whole new level.

And there's a lot to learn from stuff like this.

So that's my kinda quick look at the modern era of Debian.

Let's move into the futures that I was talking about.

So just like in cosmology... I think you all probably know where this is going to go.

You know, one of the models for the future is

that Debian is in some way going to continue to expand and grow

for however long you want to think ahead.

And there's two ways that I think this could happen.

It could be a targeted growth where we pick a direction we want Debian to move in

and we just put everything behind that

and we have enough momentum going that we can continue to maintain growth as time goes on

and meet the needs of that one area.

So we could pick, say, the server market

and say okay, we're doing all this Debian cloud stuff.

People talked about all the talks that are going to be here at Debconf about that.

There's a lot of that going on.

If you go off to any virtual VPS provider

you can pick a Debian image, pretty much on every single one of them.

It's big in that area, obviously.

Or we could say well, we're going to try to also handle desktop, or mobile, or something.

Something a little bit more targeted might be a good idea then just something that broad.

But you know, maybe if we decide, well, we just want to do this, and this

then that would help us grow.

I don't know, it's just one model.

If you look at mobile, though, and you look at where Debian is right now...

This is a screenshot of Lil' Debi, which is an Android app

that basically debootstraps Debian, that's what it's doing there in the screenshot

and this is kinda of the current state of the art of Debian on all the mobile devices

that every single person out there has in their pocket, I'm assuming

that aren't running Debian, probably?

You know, it's pretty basic,

it really doesn't give you a system that can do a lot of wonderful things,

unless you're wanting to do wonderful things at the command line

with a virtual keyboard, which isn't much fun.

You know, you can think about what we can do to expand this.

Can we, say, add Android support into Debian in some way

so that you can install Android apps and run them.

Can we have some way of getting a... you know, installing something in a chroot of this type

and then displaying it on the normal Android display

and having a full interactive application, that kind of thing.

So that's kind of an example of how we could go into one area and try to expand

to get Debian growing in that area.

The other major way that I think we could grow Debian

or that Debian could continue growing

is this more community-driven model.

This is kind of where you have different projects doing their own thing

and Debian can somehow come in and help them out.

You know, we have some good examples, like Freedombox, and TAILS, and stuff like that

that are using Debian in great ways, and doing wonderful stuff.

Hopefully they're getting a lot of developers, I hope.

I don't know if that's the case.

But there are community-driven things.

There are ways that Debian can expand out into an area without having to move the whole project there.

You can just say, it's a custom Debian distribution, it's a blend, whatever

and we're still... it's still contributing back.

It's a wonderful ecosystem going on there.

Now, if you look at something like the Raspberry Pi

I think we kinda made a mistake with the Raspberry Pi

because we said we're not going to support the specific arm instruction set that they want to use

because it's five percent faster, or something

and so they went off and built Raspbian, and that's fine, you know

but we've kind, I think, possibly, lost a little bit of the mindshare in the Raspberry Pi community

because everybody's like, "well, okay, we've got this Raspbian thing, it's not Debian, right?"

Of course it is in pretty much every important way.

And maybe if we had been a little bit more open to this project

coming and saying, we would like to build everything for armv5, or whatever it was

maybe we would have had a bit more opportunity for growth and expansion, there.

And then, if you look at just Debian developer communities in general

there's always opportunities which we sometimes don't take advantage of

to have really good relationships with various interesting projects

that might end up using Debian in some way

or might end up contributing back, or becoming part of it, even.

And so I think... I really feel pretty bullish about this community-driven thing

I think it's kinda how Debian has always worked.

I don't know if... you know, it's hard to look out and say

in ten years Debian will be an attractive target for people doing

whatever the equivalent to Raspberry Pi is in ten years

but I hope so.

So that's the one model.

Whoa, what happened to the other model?

Ah, okay, so steady state.

It's another cosmological model, obviously.

I think we could just continue sort of coasting along indefinitely

without really saying, oh, we're going to make big changes

we're going to do this, we're going to do that, we can just keep doing our thing

and be completely happy for as long as you want to look out.

We've got a lot of momentum, we can keep going.

Even if we all stop doing much today, I think Debian will keep going for years and years

quite happily

and you know, after a while, you start having to think about generational things.

When most of our generation, or generations, got involved with Debian

we kind of had some infrastructure that we just kind of thought was there

Maybe it was a kernel, or a C compiler

or something like that.

We didn't really think about it, maybe we occasionally ran into a bug in it

and we reported the bug

but it wasn't something that was at the forefront of our minds

as something new and exciting, necessarily.

And maybe that's where Debian's going

Maybe Debian becomes an infrastructure that thing get built on top of over time

and there's enough people to keep it going

because if nothing else, companies like Google, as long they continue using Debian

are going to want to employ tons of Debian developers, just to keep it going.

So this is definitely I think a likely possible future at some point

is that Debian becomes an infrastructure, and that's fine

and if you continue looking forward does it continue being infrastructure

or at some point does it get replaced

and does it even matter if it gets replaced in X years? I don't know.

But you know, I think this is another likely possibility... we'll see.

And then of course we have this final, fun possibility that you get

and I would probably have put some bullet points up here

but I had an unexpected root canal and stuff, so I kind of ran out of slides at this point.

[laughter]

You know, you can have a big crunch.

and this is always my favorite possibility for the universe as a whole.

I don't know about for Debian.

What would happen if Debian just petered and just somehow died and fell off a cliff

and everything started going down

and everybody switched to Android on their servers or who knows what.

I mean, what are they going to replace us with? I can't possibly think.

There's got to be something out there, right?

Maybe it's all Fedora in the future, I don't know.

Hi Fedora folks.

This is definitely a possibility that we have to keep in mind

and it's not like the end of the world, right?

It would only be the end of Debian, and even if that happened

think back to that earlier slide about apt establishing a new paradigm in package management.

Even if Debian stopped being actively used and developed

at some far future point that I don't want to imagine

it would still have influenced things in a great many ways

and I think we could all be quite pleased with the work that we had done on it.

Of course we all hope that it will continue to be used

for as long as long as we're involved in the project

or maybe ten years longer so we can keep using Debian systems after we retire.

So, I kind of thought that I would take a little poll of the audience.

Who thinks that we're going to somehow continue to expand

for however long you want to imagine is a long time?

Hands? Continued expansion? I would say maybe ten percent of the room.

Okay, so who's for steady state?

Slightly fewer than for expansion. Okay, big crunchers?

[laughter, applause]

Okay, well, I think we're for expansion.

So that's really all that I came here to say

It's a fairly fluff talk, I know. I hope that you've enjoyed it.

Maybe some people have some other cosmological models that they'd like to suggest?

[Audience]: Sort of relevant to the big crunch scenario

Andrew on IRC asks

I'm watching other community distributions fragment and lose focus.

Fedora, openSUSE are killing themselves right now.

Are we doing the same?

[Joey]: I don't think that we're fragmenting as such.

We've already kind of fragmented already. There was the whole Ubuntu thing

which I think is the first time I've said that word in this talk.

I don't know if we lose focus as such.

We've never really had focus, have we?

We've all just done our own thing and...

[laughter, applause]

[Audience]: I was just saying to somebody over here

that one of the differences is that

those distros are actually more tightly tied to something else that matters

whether it's the commercial distribution organization that they were sort of spawned out of, or whatever.

They've had a less completely community-driven reason to exist

and to continue to exist than Debian has

so I would not be surprised if we don't end up having an entirely different life-cycle

than something like Fedora or openSUSE.

The question I was going to to pose:

I've noticed as you have, and you made a couple of references to this

the average length of thread about almost anything has gotten a lot larger.

One of the things that I observed a while back, though

is that the average number of participants per thread had not actually increased all that much.

It was certainly for any given thread a much smaller percentage of the people currently active in the project

than used to be the case when there were thirty of us and five of us were screaming at each other.

I'm wondering if there's... I don't know exactly what to take from that

But the notion that a similar number of people can just scream at each other for a whole lot longer

and still not come to a conclusion.

I don't know if there's anything to take from that, or learn from it, or not.

[Joey]: Yeah, I don't know, I'd actually meant to say I was going to put the systemd thread on here

but despite this being a pretty zoomy thing, there are limits to floating point resolution

and eventually you can't actually represent the whole thread in Iceweasel

or whatever I'm running here.

Maybe what's happened is that we have either...

we just have more people and so the number of people who feel strongly about something

they feel much more strongly about it.

You have a small subset, who all feel that they have to win.

And so they just keep talking about this and they don't come to a consensus.

Do you have a thought?

[Audience]: It's really interesting because as a project

I think we have this sense about ourselves that we're all about freedom and so forth

and somewhere along the way freedom got translated into

"we should all be able to have our own way"

and that was really not part of the freedom that we cared about when this project was young.

Even when there were strong debates, they were debates about technical details

or when the constitution was being drafted

there were a few big questions about how should this should be structured

and then a draft got generated, a lot of folks looked at it

and went, yeah, that's close enough, and off we ran.

And the amount of bikeshedding that goes on these days just scares me a little bit

because it seems like taking that word freedom, and translating it way too much

into not needing to collaborate, or not needing to come to agreement and consensus.

Now, I don't know how we change that, or fix it.

But it bothers me sometimes when I see people take the things that I thought of

when I first joined the project in 1994

as being fundamental tenets of the project

and they use the same words, but they mean something very different.

and it causes their behaviors to be very different from what I would like to see.

[Joey]: When the constitution was originally proposed, I was kinda against it.

And I thought, well, this seems like a lot of faff around for something that shouldn't matter.

I didn't even bother to vote on it.

I was like, if Ian wants to do this, great! Ian can do this, you know? He'll take care of it.

If it breaks, he'll fix it.

And I think we've kinda...

Maybe it's just that we have a lot of people now who...

Debian is an important part of their life, maybe professionally, or personally, much more important.

How many people here in the room have their livelihood in some way connected to Debian?

So probably about as many as want Debian to continue growing.

[Audience]: One thing I just wanted to add to what Bdale was saying about the bikeshedding and stuff

I'm in preparation for my BoF later this week about the code of conduct.

I've actually been reading a lot of other codes of conduct

on a page prepared by Zack, thanks for that

and one item that I saw coming back a few times

and which I've also taken into my proposed code of conduct that we'll be discussing

is about: be collaborative.

Try to work with other people.

And I think that it could help to put something like that there.

It's just a proposal, and we still have to discuss it.

[Audience]: So let me as a dark and destructive person

focus on the big crunch model for a moment.

The question is: what would happen?

What would we be able to do in Debian if we would be in this big crunch situation?

Because, okay, now we are big.

We are very important, and we are quite central to the Free software world

in a number of ways.

What happens if this world in some ways disintegrates?

Obviously there must be a replacement.

We should be open to change and re-evolve in a way that makes the world go on

even if we in the way we are now fundamentally change.

[Joey]: You know, I didn't really think about the big crunch as affecting the Free software community as a whole.

I just assumed that was some background noise which kept everything going

even if Debian went away.

I mean, yeah, it seems to me that Debian can definitely go away

without the Free software community fragmenting or imploding

or whatever, or turning to BSD licenses

vanishing down the Apple rabbit-hole, or whatever.

[Audience]: That's not what I was about here.

It's more, we have one model of working in our Free software ecosystem

that maybe this model at some point in time is not relevant any more.

It's like, maybe some of you know this model of evolving systems.

There is a first system which is a big hack

the second system is built by a community and great and does everything

but at some point in time this second system becomes irrelevant

fundamental ideas will be changed

and a third system or third systems will evolve

on the remains of the second system.

That's just what's happening now, slowly, with X for example.

X will not be completely disintegrating

but people will evolve on it

and I think we should have some thoughts about the same ideas in Debian

and we should prepare

what might happen if this case starts growing on us.

[Joey]: Thank you for that; you're thinking further ahead than I am and that's great.

Anybody else with a question, I'm not sure how we are on time.

[Audience]: I think the disintegrating is not really an interesting point

Debian is, I think, there to... itch our scratch

if we don't have the scratch left, there's no reason to itch

as long as we are community-driven

as long there will be a scratch, we will continue to itch.

[Joey]: Or the other way round, but I take your point.

[Audience]: And to the mailing list problem

I think I see a tendency on mailing lists that we have something like

this anti-politician and anti-intellectual point.

It's too often everything that's on a mailing list that's bikeshedding

if you give a point against something

if it's not the opinion that you are, it's bikeshedding, it's not a technical argument

you are against progress

and I think we need to be a bit more collaborative at this point.

To more listen to each other, and not to dismiss everything

as everything you don't understand doesn't make sense.

It's only people that want their old stuff keeping there.

It's, I think, the reason some flames go up very much is that it's important to people

and then it's important to listen to them

and not just tell them, oh, old fart, we don't care.

[Joey]: I think if you go back and look at older threads in Debian like I did for this talk

or if you go whereever stuff's getting done, and look at what a thread looks like

when stuff is getting done and people are busy making things happen

versus when people are busy complaining about other people making things happen, or whatever

there's a really different tone there. I think you could learn to recognise that tone

I don't know if you could teach people who are part of the problem

which we all probably are from time to time

to squelch that down, or, not.

I think it's something we need... yeah. Enrico?

It's right there, go up to the stand.

[Audience]: On that point, it's interesting that you made that point

I found myself, after some frustrating discussion I was having

asking people to please... real life discussion, about something completely different

asking people... telling people, can you please stick to...

I'm more interesting in hearing your personal story.

I'm more interested in hearing your experience in what you have done.

Please don't... I'm less interested in hearing what you wish would happen.

I'm less interested in what you wish I would do.

Please let me choose what I would do, and I'm happy to hear your experience.

And I think that is a pattern that also matches very well what you mentioned.

When people are getting things done, they are not discussing about

the way they wish everybody else would believe

or the way they wish everybody else would have done something

but they bring in their experience:

When I did this last time I did it this way, and it didn't work. Let's try another way.

But when it comes from personal experience

it is more about getting things done

than about seeing who has the better ideas, or something

which is rather pointless.

So yeah, I wish on mailing lists to see people bringing in their experience

their stories at work, the way they fixed a problem like that before and how

rather than: "people should do this".

"People should do this" is possibly something I don't want to see on a mailing list any more.

[applause]

[Joey]: I think we have to somehow learn to be more accepting

of just doing something, and if it's a mistake, reverting it.

It would be great if we had more technology around this

but just socially, deciding, if somebody wants to go off and do something

then let them

and if it turns out to be a bad idea, we can undo it later.

I think if you look at where we're really good in Debian at making things happen

it is stuff like the one maintainer per package model

where people are given the power to go off and do something

and it's their responsibility, and if you have a flamewar about it

well we have processes but we don't use them very often

and it would be great if we could find more ways to expand that kind of way of doing things

off to the things which don't just touch one package.

I think that's what's broken down as it were.

We're building this bigger stuff on top of individual packages

and we don't have a way to go off and say

this guy is going to handle the systemd transition

with this group of people he's got together, or something.

Maybe that doesn't work, Bdale looks unhappy with it, so it's a bad idea

but there must be a way to make it happen.

Anybody else?

[Audience]: I used to expect that at some point sooner or later Debian would effectively just split

into multiple groups which competed with each other.

I mean I know some people talk about Ubuntu as a fork of Debian, but it's kind of a different thing

I really thought that some time there would just be a discussion

where the two sides just disagreed so badly about some issue

that you would end up with two things, basically both of which claim to be the true Debian

obviously one would probably own the trademark, but yeah, I mean

both of them would just think that they were the true continuation and hate each other forever.

That seems to have become less likely now

and it seems to me that most of the times we have big discussions

it just ends up with not much happening

rather than something happening that really annoys people.

I mean in some ways that's better and some ways that's worse.

[Joey]: That's a fascinating comment.

That doesn't fit into any of my three models, the forking off thing.

It's multiple universes, it fits into the cosmological model

Yeah, that's fascinating, why is that less likely now than it used to be?

Is there less excitement and energy around Debian or is it something else?

[Audience]: Now I would worry more that, again, if it gets harder to push new ideas

and you end up... well, we are still getting new people

but if you look at the official members of Debian, we're basically only at a replacement rate

I have to say, looking around the room, that we're definitely an aging population too.

So although that's still fine for a few decades, yeah

if we want to continue in the long term of Debian having a good future and still being relevant

then, again on your graph, how do we get back into really growing,

not just the community round the edges, helpers and contributors and so on

but people who are members of Debian should also be growing

and taking new ideas.

[Audience]: Sort of replying to that: if we go a bit smaller than cosmological, and go to galactic, say

I think Debian could be looked at as if it started out being a star nursery

and then we turned into a galaxy

and we're now at a stage where we need to find a way of maintaining the black hole

because otherwise, if people aren't allowed to work on an alternative black hole

then the arms will fly off, as... yeah, we need to suck more.

[laughter, applause]

So the black hole is the sort of boring, central packages

which you're not allowed to touch, because if you do that everything will break

and we need a way of instantiating a new galaxy next door

and just replacing the black hole, and as you say if it doesn't work

you can git revert.

So, and the other thing is, if you look at the mailing lists

you get the impression that there's a war going on

where there is going to be a schism.

Half the people will go off and maintain their servers

and the other half will go off with their tablets, or whatever, and sort them out.

But actually, the people in those discussions aren't going to build either of those things

and the rest of Debian is just getting on with it.

So, that's why I think Debian doesn't fragment

because the vocal people aren't necessarily the people doing the job.

[Audience]: I think there's another possibility

and that is that when I think about Moray's question

There are more derivatives of Debian than any other core distribution

so there are certainly lots of people out there who have decided

that the thing they wanted to to differently, or cared about

was worth going, creating a CDD, or a fork, or whatever.

So that's happened, it just hasn't dragged the trademark into... or the name into some kind of a pit

which I would hate to see happen.

But I have this sense that maybe the other thing about it is that Debian has become large enough

and means enough things to enough people

that the vast majority of us in the project who don't give a flying you-know-what

about whether it's upstart or systemd

That's an impassioned important discussion for the people for whom how the system boots

is the thing they care about in Debian.

But for the vast majority of us it's like, as you say

I do that once per kernel update cycle, a reboot

and the rest of the time I just don't care

and so the idea that the distribution would fracture

or somehow Debian wouldn't be Debian any more because there's a fracturous discussion

going on in a particular sub-project or sub-part of the distribution

is just hard for me to wrap my brain around.

[Joey]: It seems like it would have to be something that isn't technological based.

Some kind of, you know, we want to change the social contract

or maybe want to change what free software is

and that would fracture Debian.

[Audience]: So, on the lines of what Bdale just said

this way that we are becoming almost a preferred choice to be upstream

is a very good thing

and that enables our work to scale much better

than if we try to grow the project

and I think the reason why we aren't growing in terms of number of people

is that we're already at some kind of limits of scaling

We're having... a lot of things we're talking about

are difficulties to do with coordinating and communicating between this number of people

and allowing, and becoming upstream for people is a way for us to scale that a lot better

and one of the things that we should be trying to do is to look outward

rather than inward

and to try to think of ways in which we can be a better upstream for people

to make it easier for people to derive, so that fewer people have to do their work within Debian

and that they're easier to do it outside Debian.

Because after all, software freedom is about freedom to make the change yourself to the software you're using

and that doesn't necessarily mean that you want to have a huge, kind of

get involved with a huge complicated upstream who have processes

and decide to do things a particular way.

No, you should just be able to it.

At the moment if you want to do that it's quite hard, and we should make it easier.

[Joey]: Yeah, you know, when you think about that

maybe it's kind of what's happening now, but you have to wonder

your scenario, you can go either one of two ways.

You can have a lot of custom Debian distributions, and things based on Debian

and Debian can just become a background infrastructure

and then who wants to work on it when it's some thing that's down there in the depths

that other exciting things are being built on top of

You know, maybe you contribute patches back when it makes your life easier

but do we get a sustaining model that way, or maybe we don't.

I kind of used to have this argument with Manoj.

I thought that Debian had to expand or we were just going to die

and Manoj was like no, Debian is just about what I need for my system

and what my friends need for their systems.

I'm only interested in it in that way.

And I don't know, maybe Manoj was right.

I think that I was definitely wrong.

The best arguments are always that way, right?

[Introducer]: Okay, the time is over, so we have to take this as the closing comment

and, yeah, you have to move it to lunch to discuss over that.

[Joey]: Okay, thanks everybody.

[applause]