994_Debian_Contributors_BOF.ogv

Good afternoon everyone

So, Enrico is a guy with many hats in Debian and

I didn't expect it but he will talk about a new hat:

the "Debian contributor hat"

So, please welcome Enrico.

[applause]

Nobody expects Debian contributors.

Our chief weapons are Debian Developers.

Debian Developers and Debian Maintainers are our two weapons

Debian Developers and Debian Maintainers and Alioth accounts

are our three weapons

and Debian Developers, Debian Maintainers, Alioth accounts and bug reports

are our four amongst our weapons

amongst our weaponry

are such elements as Debian Developers, Debian Maintainers are... I'll come in again.

[comment about the Social Contract]

Nobody expects Debian contributors

amongst our weaponry

are such diverse elements as

Debian Developers, Debian Maintainers

Alioth accounts, bug reporters

and it miss contributors

There's more possible

We clearly need another hat

to make sense of all this.

Right?

so let's make a new hat

I would like to create a hat

of Debian contributors

so that we can say

Nobody expects Debian

our chief weapons are Debian contributors, full stop.

A hat for everyone that contributes to Debian.

Now, the details are here.

This is the file

with notes.

Is it readable at the end?

No?

You have much better eyesight that

No, sorry, I open in a different terminal.

[laughter]

Right.

This is a text file with notes about what I plan to do

I would like to go through it with you

and then

edit it together with you

and at the end of these 45 minutes

have a proposal that

I'll try to implement during DebConf

and fail unless somebody helps me, probably

but what we see how much of this

we want to do, how much of it

can be implemented during DebConf and

at the end of DebConf

I'll make another talk about what we changed in Debian.

So

This is probably a good time to

Ok, so

the problem we want to solve

the problem I want to solve

and you probably also want to solve

once you realise you have a problem

is... words wrapping.

[laughers] Now that we solved word wrapping.

Ok, so

in Debian, we have

lots of contributions that

we do not thank people for

in the sense that we do not acknowledge

that people that did

translations,

worked

we don't acknowledge somewhere people who did translations

or people who reported bugs

There is lots of contributions outside of Debian

but if we look

around package tracking system

Debian developer package overview

it's all package-based

because that's where we come from

and

and so we track uploads that are easy to track

and we acknowledge uploads mostly

but

there's so much more that we do not acknowledge at the moment

Also,

There is a need

Over the time, we created so many ways of contributing to Debian

outside being a Debian Developer

so

the head

of the Italian translations in Debian

is not a Debian Developer

Nobody knows

who she is

Well, I know who she is

Francesca told me

and at least two Debian Developers know who she is

I've no idea who's the head of the

Vietnamese translation

Raise your head if you know who is the head of the Vietnamese translation

in Debian.

[chat in the public]

[public] I know the French one

yeah, we all know who the French one is.

[laughter]

[claps]

But, we totally need a way to credit people

properly.

Debian is so much more than

uploads and Cheese and Wine BoFs

So, there's this idea of creating a hat

of Debian contributor which is

the ultimate do-ocracy hat

There is no bureaucracy involved

If you contribute to Debian

you are a Debian contributor

automatically: there is no new Debian contributor process

If you stop contributing to Debian, you stop being a Debian contributor

That's it.

When you are a Debian contributor

you get such

diverse privileges as

having your name on a list

of Debian contributors

[laughter]

and possibly, if you're not a Debian contributor anymore

then your name will be on a list of Debian contributors in 2012.

If we want to be a bit fancy, when you click

on a name on a list

you get a page about that person

that list what they've done

And that's about it.

That's all I had in mind

You report a bug

you become a Debian contributor

Then maybe everything you do for Debian is reporting a bug

on the 11th of August 2013

then you will be a Debian contributor

from the 11th of August 2013

to the 11th of August 2013

at least you're acknowledged

as that you've helped Debian

Mostly I've said all of this

In order to do this

there needs to be a way of collecting this information

There are data sources we can easily tap into

such as

package uploads

but we can see mailing list traffic, easily

We can get

bug tracking system activity hopefully easily

We can hook into

commit logs of Alioth

version control systems

[phone ring]

[phone's owner] Sorry

All right

At least I don't have to answer

And, this system, whenever

at any point of its usage

would be unfair

because there will always be some bit of Debian that

isn't yet providing data

for the list.

However, the point is that by having such a system in place

we create

a proper way of acknoledging people's work

and we give motivations for teams to contribute

contribution information to the list

Questions so far?

[Q] If you do this automatically

which we can here assume you want to

wouldn't that raise privacy concerns?

Sorry, I did not understand the second part

[Q] Wouldn't this raise privacy concerns?

Good question.

The question is,

if we collect this data automatically, wouldn't there be privacy concerns?

Yes, possibly

To be discussed.

All the information presented here

is already public

There is a public record of package uploads,

mailing lists are publicly archived

but

the BTS is fully in the open

commit logs on Alioth are public

I would only mine
commit logs on Alioth are public

I would only mine

or get data from sources that are either public

or that people agree

to disclose

However, it's the usual issue of

the information is public but not presented in that way

for example

when

when

I collected information from Debian changelogs to

see the history of people's contribution in the process

I did not make that information publicly available

because it's a bit

I would like to have a discussion

whether we want to present that information in

collated in such a way

So yes

that is an issue

My general idea would be that

once the system goes public

there should be a way for people to opt-out

of having information about them displayed

Most of the privacy scenario...

so, generally people

I would understand the general...

The default would be that somebody likes

their work to be credited

However, there may be people

that

would have issues if some of their work is credited

Most things revolve around

job hunting like recruiters looking at the internet

and you either want to be found

or if you don't want to be found

you already have a problem when contributing to Debian.

I need to be careful about contributing with a different name or something.

But, yes

my assumption is that

It's not a big privacy issue to get it started

as long as

we give people the opportunity to opt-out

Does it seem reasonable?

[Q] Maybe with a opt-in

Maybe with a opt-in

By default it's not opt-in

I would be fine with opt-in

I wonder how many people would know

that they can opt-in and how to opt-in

At the moment

the current Debian infrastructure is not even opt-out

You can't say "Please don't show

the messages that I sent to Debian mailing lists"

I would see surprising that such a thing would be opt-in

because anything is already out by default

On the other hand

I could be convinced

about making it opt-in

It's seems to be

at the moment something that would

impact

the start of such project

So, before

somebody gets a useful list

I would need to go and bother a thousand people

to opt-in the system so that at least can get populated

Yes?

[Q] I think you should also look at legal

aspects, you are collecting personal information

as far... I don't know the law

by heart

but as far as I understand it

it would be illegal

in half of european countries

Then Debian is already illegal

No! Currently

you are not collecting this information

Yes, we are

You have it in different pieces

and at different places

and it start

The package tracking system would be illegal

No it's not illegal

You don't collect information there

collated about a person

There, you have it

to follow specifics purpose

and this is not illegal, but if you collect it

and collect all information for a specific

person or all persons

and collate it

this could possibly be illegal as far as I understand it

I would like

You said that you are not sure

I would like to talk with somebody who is sure

because we already have things

that

would potentially be illegal in that view

like the package tracking system

goes beyond

listing of some of the uploaders

because it collects information

about everything visible about the person, bugs and so on

[Q] Would it be

a preset that is useful to

structure this kind of things by, I mean

some

reasonable set of people

listed as contributors

You could start by assuming that

people who have said

"I would like to be a Debian Developer"

or "I would like to be a Debian member"

have implicitely given their authorisation

to be listed as Debian contributors.

It's a subset

that we might have to run up by a lawyer

but it seems a bit more reasonable than

??????

I was just gonna say I think

the legal thing should probably be a separated BoF

because

in the practical terms if you're only worried

from the legal point of view, you just need to

get someone who is in Debian in the US

to host a US server because the US

doesn't basically have this kind of laws

It is rather missing the point.

Exactly, the relevant question is I think

no about the legal thing it's a kind of ???

but about what we want to do and what we think is fair and so on

Well I keep thinking ??? something

much more lightweight than launchpad

is doing

Another point of, another way to mailing list would be to

do this automatism

without publishing the data but

with afterwards sending

about once a year or

even possibly less

a mail to every contributor if he wants to be listed

so we have an opt-in instead of an opt-out

If we do it seldomly enough

it would not even be so

stressing people

Yeah, that makes sense

You don't need to wait a year,

you can do it the first time you detect someone new

someone starts contributing

we can send "hello, you did something in Debian

we would like to credit you"

Wait for a month.

Wait for a month after the first contribution

if you get more, then

say "Do you want to be thanked?"

Makes sense

[Q] Even without waiting. We can do as the Linux kernel

you knows the stats about

which companies contribute to the Linux kernel?

It's like that, they check for new commits

and just send an email to ask affiliation

There's other systems that calculate karma

Those are illegal because

then they produce

information that could be used to rank people

from a job point of view

[from audience] there is Google

The other... yeah, well.

I'm not that concerned with that regard at the moment

[Q] Would that credit

only be for

persons, or also for organisations or companies?

I'm

interested about

entities

directly contributing to Debian

So, if

there's some email and GPG key

which belongs to an entity called

FooBar ltd

then FooBar ltd will be credited

I'm not interested

that there is a real person behind this

and not interested

that's the real name of a person

It's just whatever

is chosen for contribution

that we only check identities

when we have to

which is when we give people upload rights

because we want to know where they live

and go there to tickle them

[laughter]

I think that

the main index would be the e-mail

Even behind the scenes... But the main index would be the email?

That is one issue

That is an issue

because depending on the data sources

the index could be an email

in the BTS for example

an Alioth

account

for commits

in Alioth

or

Debian Developer login name

for sponsoring uploads

so

possibly, that

that could be some

need of mixing things at some point

which is a bit of a separate problem

I don't mind starting

with that not being perfect

or being fixed by hand

And if we evolve at some point

there may be a way for people

to log in and prove that

they control that GPG key

or that e-mail address or that Alioth account

you know, sending e-mail challenge saying "that's my e-mail"

but ok, it is

yes, ok, then I merge

the two sets of contributions

possibly for next year

because it starts being a bit heavy weight

but there's a problem of doing this collation

it is a wider something that

greatly help the MIA team

So this

in the end can be something

used to detect when people

are not contributing anymore

I would also

want to make it so that

if you are a Debian Developer

you are not automatically

a Debian contributor

[laughter]

If you are not a Debian Developer

you are not automatically a Debian contributor

you are a Debian contributor only

if you contribute to Debian

If you are a Debian Developer that

does not contribute to Debian

then you are not on that list

You keep being a Debian Developer

There's the usual

I'm not interested in changing the rules

for

getting removed from

Debian keyring

I think what we have is fine.

But

it's perfectly right if

if I'm not active in Debian for a year

but I still care about the project

involved, and follow things

but I don't contribute anymore

then I'm perfectly happy that I'm not listed

among the contributors for that year

There was a hand?

[Q] Yes, I'm relaying a comment from IRC

Wouldn't

recognising every small thing like

just one bug report

dillute the

the weight

of a Debian contributor

that so

are there Debian contributors like

the head of a translation team

be equal to

one bug report

That's the comment, I'm not saying that I agree

The way I see it

again, we can

you can tell me if you don't like this

but the way I see it, they are the same

They are Debian contributors

maybe one has been a Debian contributor

for much longer than the other

so the time span

of the contribution is the only extra

bit. Or you can click on a

a person and list their contributions

But, I don't want to

calculate a number

of how much one is a contributor

I don't see

why one would want to

rank people

by contribution

I'm not interested in that

I think it creates a kind of

community where I don't want to be in

personally, I mean, that would be my personal feeling

But I would be able to

say "thank you" even if

a person reports a bug

It takes time to report a bug

There was an idea

talking about this

informally earlier

of

having some data sources used

only to compute

the time span but not

the status of contributors

for example mailing list traffic

Maybe people don't like the idea

that if you send a e-mail to

a mailing list you are Debian contributors

There's indeed many people

who send lots of e-mails to Debian

mailing lists

but do not contribute

[Laughter] to Debian

at all

However, if somebody

is a contributor for some other reason

then it makes sense to

look at mailing list involvement

say: "Well, you reported a bug today

but you've been active in mailing lists

for a year"

so you are a Debian contributor because

you reported a bug

and you've been a Debian contributor for a year

because

you've been into Debian for a long time

so splitting

the...

I think some data sources

are only useful computing the time

and not the status

I would like to add something about

the issue with detecting the people

because we had some experiences in the team maintenance thing

and, in UDD you have

this Carnivore database

which is based on the

key fingerprint

and we even have

for these people

somebody with five

different names

names spellings and so on

people are using different e-mail addresses so

I really really really doubt

you can manage this for

say, more than 500 people

because it's

manual work and we tried it

it's hard to cope with this

No no, I don't want to do manual work

Yeah but, ok

Good luck with automatic detection

No, I don't want to do

that much automatic detection either

I would

like

things to be fixed as much as possible

and

at lower level of data sources

When that is not possible

I would like to offer people a way to

fix the data for themselves

There's that

unpronounceable and

evil web site that tracks

free software developers

that in my opinion should be illegal but isn't

it starts with "o", ohlo?

"O-l-o-h-o" something like that

I actually had an argument with

their CEO, saying

I've asked to opt out

of their system and they told me to fuck off

So

talking about privacy of these things

I guess if they can do what they do

I think we don't have a problem to say "thank you"

to a bunch of people

They do of

they have a mess, their dataset is a

a mess

but people still like them

I'm listed about

twenty times in their system

and if I really care about

contributing them my identity for free

for them to send it out

then I

can log in and

merge these identities for them

I guess

that's something we can offer

and

and that can improve things

like carnivore and so on

and possibly we can have

personal homepage...

for contributors

[Q] Where they actively

contribute to their identities?

At least we've done

all we can to thank them

but

Well, you know, if I tell you "thank you"

I don't expect you to say

"You didn't thank me in a an appropriate way"

Right?

On the other hand

I'm ok if people say

"Well, I'm also that person

you don't need to thank me so many times"

and that

that's fine, but, on the other hand

I'm perfectly happy if

people want

to use multiple identities

for contributing to Debian

Somebody in Debian science

may want to contribute to Debian games under a different name

For similar reasons, I don't want to trunk

the time frame of contributions

below the month level

I don't want

to say

"You have contributed to Debian games

between

10am to 11:30am on a Thursday

when you were at a meeting

[laughter]

Right?

So, it's

I want it to be coarse

It really needs to be mostly about thank you

and about building reputation

Reputation is also what we can reward people with

which is kind of the point of the exercise

and that reputation is nice to acknowledge

and at the same time it's nice

that we can

look up someone's reputation

at least

as a Debian account manager and Front Desk member

if I can see what's somebody's reputation and have a

reasonable look

then I can make the process much swifter for them

and especially for non uploading DDs

it's currently very hard

because

I can go and look at package changelogs

to see that they have been active for ten years in the project

and

if somebody is a non uploading DD

that has done translations at the moment

I have very hard time seeing

what they've done

possibly I don't understand the language

So, well

and that makes

makes it easier and

that another

outcome that I would really want to see

out of this is that

we start to actually see

in front of us

that Debian is not just

about technical development

There's a lot more

but if people still perceive

"Oh, I'd like to contribute to Debian but

but I'm not a technical person."

and

many of us may have

difficulty in pointing out places

but one can go and have a look at

such a list of people and say: "hey, look

there's people over there that do also other stuff

it could be indexed by

topic of contribution

as well

Again

uploading, these are contributors that do uploading

these are contributors that do translations

That may be introduced at some point

so that we

we can turn on a spotlight

in several aspects of

Debian that we usually don't look into

we just take it for granted

[Q] I have a lot of questions

Yeah?

I'm more concerned, more than anything

If this...

I mean, some

One thing that you said at the beginning is that

maybe some people would be quite interested

in this for collaborate research, for work

for creating a curriculum

[A] Yes

[Q] and

if it's as easy as to filing a bug which

can be very hard to do

depending on the bug or very easy to do

depending on how much you put in it,

wouldn't be able to

to calling people to put this

random stuff or easy stuff there

for just being in the list

Yes, well

at that point they're really bad job recruiters

Say that if there's a recruiter that just

does random google search

and contacts whoever comes out

like Google recruiters

I keep being contacted for things like managing clusters

If anyone's had a look at what I do

I don't manage clusters, right?

So, if there's recruiters like that

there's nothing you can do

If people care to actually go in and to look

at what a person has done

then

there's a way to find out

what contributions actually were

from just linking to

the usual Debian mailing list

archives or uploads or

commit logs and so on

[Q] So, if I understand correctly, you are ready to

use the BTS as a source

for the contributors, and you could have

a lot of nicknames there

It could be "the Pope" or "Bill Gates", or

the name of a serial killer or whatever

Are you ready to use that?

Maybe there could be some issue with

people that would be falsely

credited

[A] We try and see what happens

It could be that for the bug tracking system

we need to have

some extra intelligence built in to avoid crediting

China manufacturing corporations

or viagra supplies for cheap

mailing list archives are being spammed

somehow

but again

then, maybe

maybe there's lots of noise

and we want to set a threshold that one

contribution to the bug tracking system is not good to be enough

or maybe we can look if that bug was closed

because we close spam bugs

anyway

If there's cruft, if there's noise

that we can filter somehow

maybe we'll loose some contributions

but at least it wasn't for lack of trying

I mean

I don't want to go out of my way to thank someone

If I need to

find out where you live

to come to your home and say thank you

because I don't see you anymore

or because I met you in a crowded

square and you gave me directions

but I didn't take down your phone number

then, it's unfair

you won't be thanked

when I finally reach my destination, but

that's life

So

then maybe it takes

ten e-mails to the BTS

with different content

to be acknowledged

It's possibly something to be

I would leave freedom to the people doing an import

thing, something that gets data out of the bug tracking system

give that area freedom to work it out

and I don't want anyone to be perfect

The system is not going to be perfect

if

somebody would really like to be credited but

isn't in the list

and what they do is report bugs

I would be surprised because if

if somebody wants to be credited on reporting one bug

I'm going to say " Well, just upload some more"

I'm not sure I want to have people

in Debian

that contribute

in order to be credited

If

If contributing some more is

a way to get credited, then

yeah, we don't need to be perfect

just ask people to contribute some more

or if people contribute

the problem is when people contribute a lot

but they're not credited

then maybe we need to figure out how to import

data from another part of Debian that

we currently don't track so well

I'm totally not looking for perfection here

[Q] Have you thought about other ways to use mailing lists

archive because you said

we don't always want to use it as

source of contribution but more

to track the length of contribution

but there are some lists, for example

l10n translation lists

where I guess most people who contribute are real contributors

and sometimes they

also use

some code

I mean, the e-mail could be detecting

like "Request for review"

and stuff like this so you can add some hints on the

type of contribution more precise that

just "sent a mail to the mailing list"

and maybe also for a user mailing list

not tracking the question but only answers

or sort like that...

because, well, asking a question is not a real

contribution to Debian, but helping someone else to

use the Debian system is

a Debian contribution

and one of the most important one to start with I guess

[A] Decoding mailing lists patterns

fairly is not

easy as far as I understand

but I guess it changes from list to list

but

that could be solved by

adding a general mailing list

source that only looks at time frame

and having the translation team saying

"In our mailing list

there's more semantic information that can be

extracted"

so make

an extra import if

importer for my team that decodes

things and put there properly

as long as you can have

many things dumping data into a bucket

then that works

I currently have no idea

well, only vague ideas on

how...

what's the protocol to collect this data

I would

from the central bit

I wouldn't go much further than

having a list of urls from where I download something

on a regular basis

because that's the simplest

protocol to identify instead of

submits and whatnot

and what's in the file that gets downloaded

could be identity, initial time and final time

and then each datasource

makes a file available somewhere

I was thinking of something

super simple minded like that but then it's not enough

because then if you want to look

to make a link

showing the contributions then

well I need something

about how to

how to make this data available for collection

at the central point I'm only

I want it as simple minded

as possible, I don't want

the core of this to be about data collection

about mining

it should just be about collecting what as been mied

One last question?

[Q] Just a thought on what you said there is a

Google Summer of Code project

on Fedmsg

which is a ??? maybe to

have a bus

collecting lots of information from

various sources of information

among those, there's already

mentors.debian.net

and I guess the BTS would be

[A] main source? [Q] it's a source

???

yeah but I guess it could have enough metadata

like ???

the software is fedmsg

F-E-D-M-S-G

it's from Fedora and

the summer of code student is trying to adapt it for Debian

just a software bus

where you get lots of events from various sources

maybe it could be

assuming it goes further than this summer

it could be an interesting use of this project

Ok

[Chairman] As the mentor of this project, I agree

[laughter]

Ah, let's talk

and I think the student is arriving today, so

[laughter]

So, we're out of time

so, thank you Enrico

[Enrico] We can talk about this, I plan to work on it during the DebConf

shortly, that should be Martin Ferrari coming who was also

intending to work on a site, similar to this

and I get we should all get together

I can think of something

and feel free to stop me to talk about this and

offer help or

that's the main thing I want to do a DebConf

and I want to get something land before

the end of DebConf so we can

present it properly

[applause]