GitHub Git Merge 2013, 23 May 2013

Previous Meeting

⏯

youtube image

►

From YouTube: Git Merge • Google Summer of Code (Git Core Team)

Description

Subscribe: http://bit.ly/subgithub • A wide-ranging discussion about the git project's involvement with Google Summer of Code, featuring Jeff King and Thomas Rast as well as others.

About Git Merge
Git Merge is a Git user conference held to bring together Git users and Git developers. This May, they were brought together in Berlin to improve the world of Git for everyone.

A

Ok, so,, who wants to talk about summer of code stuff?, I think maybe everybody does to some degree,, because I think libgit2 people were a little disappointed that they didn't.

A

Get to have a student that was...

A

and that's very sad,, but in past years we've just kind of combined them.

A

So, I'll sort of start with what happened with summer of code, this year,.

A

Which is, basically, we're..., I think thomas started us off.

B

In maybe february or january saying, "hey, do we want to do it...

A

You know,, we need to come up with ideas," and the ideas page kind of filled up mostly with projects.

A

That never got picked in previous years,. A couple of new projects.

A

And then, I kind of came about a week before our application was due and said,.

A

"Hey, you know, these projects, suck".

C

And I don't mean they're bad projects,. I actually think they're things.

A

That would be cool to get done,, but I think they're projects that:.

C

A), they didn't get picked in previous years, and there are some new projects. Too,.

C

But,, so I think there's a good chance they're not going.

C

To get picked again this year, but b), I think the scoping is wrong.

A

That we have problems of saying, "I am someone who works on git.

C

And I know a lot about git, and I know this is something I would find really cool if we could figure out,. It's a long-standing problem with git,, something like dealing with large 'blobs'.

D

I get log-l work that thomas worked on,.

A

These sorts of things.- these are awesome, projects, and we get excited about them, and they're the things we want to put on the ideas page because they're awesome., but it also means they're horrible projects for somebody who's just jumping into the code base to start on.

A

It's one thing to say,: "these people can..., you know,.

A

They have three months of summer of code to work on this,.

A

That's a long time,. I just contribute in my spare time,.

A

That's a lot more time than I have." but,. You have to bear in mind too, that they're a lot less experienced both with git,.

A

And they're also a lot less experienced in general,, because they're ususally a lot younger, and college students, and things like that.

A

So,, I think one of the problems we've had, and I think we've had a lot of successful.

A

Gsoc projects, and I think we want to.

A

Be a part of the program again, because I do think it's good.

A

But one of the problems we've had, is that we have these really cool,.

A

Big ideas,, like a great example, is the rev-caching work.

A

Done by nick aidalin in 2009, 2010,, I can't remember now.

A

And it was really great, ambitious project to say,. Let's cache incremental traversals.

A

Of the revision graph in a way that we can reuse.

A

Work for things like upload, pack, or pack objects, and things like that.

A

But it was such a big project, that at the end he basically...

A

The gsoc length of time was enough for him to come forward and say,.

A

I've got a thing that kind of works,, but nobody really understood it very well,. It wasn't integrated into the code base very well,.

A

We weren't even sure if it was the right approach or not,.

A

And so, of course, at the end of the summer he went back to school.

A

And was very busy, and I don't blame him for not working on it. More,.

A

It was a huge project and he had lots of other things going on, but in the end, it didn't actually help the git project at all.

A

Because it never got merged., so I think I'd much rather have smaller projects.

A

That are less about doing something amazing for git,.

A

Something that we're like,, "oh, man,, how come we haven't done this in the last five years", well,, because it's hard, and doing a project that is more about getting the student.

A

To integrate it into the community.

A

having them work on a smaller project that is less glamorous,, but that will actually get done, will actually follow the approach.

A

That we take with our own patch series,, where you actually,.

A

My average patch series,, if it's going on for more than a month,.

A

It usually means that I need to be breaking it up into smaller pieces. That, nobody's going to review your hundred patch series,, because that's just awful.

A

so,, my feeling was,. We shouldn't do summer of code.

A

Ir we're not going to come up with projects that are moving in that direction,.

A

And then I think there was a lot of dawdling and a little bit.

A

Of discussion on the list and then the deadline passed,.

A

And we said, "oh, well,, it's probably better to take a year off anyway.".

A

So,, that's what happened with summer of code..

A

What's going to happen next year?, I think we should do it, and I think we should follow the direction.

A

That I just outlined, but I also don't want to make a decision like that unilaterally,.

A

So,, what do other people think, do? We agree. That's a good direction to go?.

A

To me, the two issues are project selection, or project proposals having a better scope,.

A

And also integrating students into our community a little bit more.

A

And I think libgit2 has actually done a lot better job with that,, because.

A

You've had projects,, I know carlos,, you were a student,.

A

Dissent was actually a libgit2 student originally,.

A

And [chu] was a student. This past year,.

A

And I think that's worked a little bit better,, I don't know,. Maybe you choose better scoping,. Maybe you just are better at getting people integrated.

A

So anyway. those are the directions. I think we want to go,.

A

So I'd like to hear discussion from people on, is that a good direction for us to go,? Would that solve some of the issues? We've had, and I think thomas is maybe going to give us some numbers,.

A

But then also,, if we're going to do smaller, projects,.

A

What kinds of smaller projects?? That was the thing I think that was the blocker. This year,.

A

Nobody said okay, well, here's we need smaller projects,.

A

Okay, well, here's the list of five smaller projects.

A

and that part didn't happen. and I think it would have happened,. We just kind of didn't start talking about it until too close to the deadline.

A

To do those things, so., I could just keep talking while they get ready.

A

Do you have ideas for small projects?

A

to be honest, no.

A

(laughs)? I mean, there are a few.

A

And I think one of the things that makes me think that's the direction.

A

We need to go, is matthew, moi, who's, a professor of something.

A

At some place in france, I don't know, but he's a frequent contributor.

A

And he has his,, it's insemag, is the university,.

A

And he has his students,, I guess he runs a summer program.

A

Where they take a class from him- and he puts them in maybe groups of four, and has the group of four work on a project that is,.

A

Honestly,, the scope of it is not very big at all..

A

They end up being these things that you're like, "I could probably do that myself in about a week.", and he has a group of four students. Do it in about three weeks.

A

Which seems crazy,, but it's about the right time, for somebody.

A

Because it's not the project, that's the hard part,, it's figuring out how open source, works, and figuring out how the code works.

A

And getting integrated into the community and all those things. so,, I think we've had a lot of really good success with that, because he's figured out how to do this scoping.

A

Because he gets on his students, to, he's like,.

A

I don't know how he does it,. Maybe he promises to withhold.

A

A good grade from them if they don't participate in the git mailing list or something, but, presumably,, that's what he does and they end up actually becoming better integrated with the community.

A

And I think most of them don't stay,, but I think they've learned a lot more about how to do that. Interaction.

A

And their code gets merged,, so everyone's happy.

A

So who chooses those projects? uh,? He does., I mean,. He basically runs his own summer of code, himself.

A

Do you know if he does intensive mentoring of the students?.

A

I have seen him do mentoring of them on the list,.

A

Like he's cc'd me on all the patches, and he helps with the review.

A

And makes sure that gets pushed along,.

A

I don't know what kind of mentoring he's doing.

A

Offlist., I assume there's some., I might have misunderstood the program,, but I think.

E

They're in a room together working on the patches.

A

That could be, yeah.

E

Like their first contact is like, a patch with multiple sign-offs --.

A

Well, the groups are,, but I don't know if matthew is working with them or not.

A

Yeah.- and they certainly have each other. and that's something that gsoc actually discourages.

A

you're not supposed to work with someone else.

A

I shouldn't say: discourages,: they want you to become.

A

Part of the community, but what they don't want is for students sharing a slot kind of thing,.

A

Because it just creates too many headaches.. Are you ready?.

F

Uh... no., I just wanted to add that matthew's projects are, I think,.

F

In the hot phase, only three weeks, right?.

D

Yeah,, I think that's--, I'm sure, there's some preparation,, but three weeks.

F

Is far smaller than the average gsoc,, which is about ten or something?.

A

It is,, but you're also talking about several students working on one three week: project.

A

Versus one student working on a,, I mean, because, as we all know, manpower is shiftable between time and people., I actually think the idea of a 3 week, project, though,.

A

Is interesting. I just think about, in past jobs that I've had,.

G

Where we've brought in interns,, it actually is incredibly helpful.

G

For an intern to start out with something that is a one to two week, project.

A

And maybe shepherd them like that...

C

Yeah. and as you pass through, you learn the process--.

G

A

I mean, to the extent that we could, and I don't know, what's possible under the.

G

Terms of google summer of code,, but to the extent that we could say this is a two-phase project.

A

Oh, yeah,, we can totally do that.

G

You're gonna have this initial set of things that you're going to do where we might judge it literally like a weekend's worth of work.

G

But to give them, make sure that there's a shippable--.

G

-- yeah, and I think that's the key part,- that there be a shippable phase,.

A

Because that's what we run, into, that there's not...

A

The first important milestone is like, one week before the end of the program,.

A

And then we don't, they never ship anything kind of thing.

A

Even if it's something we consider trivial.

A

no,, it should be something we consider trivial,, because then it's easier to mentor them on the community issues and the integration issues. yeah.

A

I think the easy parts should be done before the summer of code...

A

Yes, some-- --, that's what I did when I did the... with my students,. I consider it successful.

H

In the way that stuff got merged into master.

H

And most of the work he started ended up. Neutral,.

A

And it was less successful.

A

right. well,, that's not always going to work. Out.

A

I mean,, I don't think that--.

H

-- should be on. The first parts should be the first patches before you get accepted.

H

So that people can provide or can choose their proof,.

H

They can use email, and stuff like that.

H

I had one student who actually said, "I, really want to work with you.". I said "okay,, here's, this really easy task.". It was pretty easy.

H

I never heard from him., he didn't do it.

A

Right. so yeah,. I know that a lot of other summer of code.

A

Organizations do that,; basically they won't even take a student until.

A

They've gone through that shipping thing.

A

And that they stand up to what they say.

H

Yes., the next thing is that I co-mentored for thomas with a git lock-l option.

H

And I had the impression that it was a huge task, and both didn't finish it and it didn't get much of an advanced state.

H

Because it was so huge., so I decided to break it into small parts.

H

That became suddenly aware, so we have this little task, less,, which will last one week or so,.

H

Or a few weeks, and he could push one patch into the list.

H

And start working on the other thing,, so there was no dead time.

H

What you have when working on just one series.

H

So, when he was waiting for class,, he could work on the other stuff.

H

And it was surprising that I really think I could have done the same thing.

H

In much less time than I was planning on mentoring, him,.

H

But I think that's the point. right, so yeah,. I think that's a mistake that we certainly made early on.

A

Was to think "oh,, it's like free pony work," and it's not.

C

It actually takes more time to mentor someone to do the project.

A

Than for you to do it, yourself.

H

But, hopefully, you're--.

H

--, you have to do it for different reasons: right, exactly. and I think that worked out really well,, but well, students are different.

H

Sometimes you get a really bright one,.

H

Sometimes you get carlos.

C

Then you can even extend the scope of the gsoc project,.

H

You have to do this,, otherwise I'm going to fail, and that's what I think you're dealing with.

H

So,, that's what I propose on that topic,, but I think personally, that we could use a [inaudible].

H

Right., but it's one piece or one area of it: you just have to be,.

H

Have to get used to in a limited time, and we can put in other merge drivers.

H

So yeah,, I would agree with all that, and I think a lot of it comes down.

F

To the simple idea, of, get them to work, how we work,.

A

Because there's a disconnect., we don't go away from the list.

A

And come back three months later, with a giant, with something,, we have lots of little projects going on, and that's something that we haven't.

A

So it may not even be, change the scope of the projects,, it may be, break them down. Better.

A

And maybe, choose topics which are on the border of git and not in the core.

A

Yeah,, I also think that is a problem, yeah.

H

Still some kind of topical core,, so they are easier to tackle than rewriting.

H

I actually think, related to that,.

G

That libgit2 has been more successful,. Perhaps, is that everything is still on the border.

G

(laughter from audience).

F

So yeah, I made some statistics.. I have this slide also,, but I think lunch will do for that.

F

I did this yesterday on the train, so about four hours in total I think,.

F

So it might be very wrong,, but I just wanted to dispel this impression.

F

At least I got from the thread, that everyone has the impression that we suck at google summer of code.

F

As far as I can see,, this is absolutely not true.

F

So,, it's very hard to come by other projects, or other organizations.

F

In gsoc terminology statistics,, especially if you're on a train and 3g internet,, if at all.

F

Ourselves, you can dispute these assessments,. This is what I could figure out in the time. Given,.

F

And I should also say that the [inaudible] who's, unfortunately not here.

F

Also contributed some corrections to this.

F

We have slightly over half,, you see at the bottom. It's not the same. Twelve,.

F

But slightly over half of the projects that I estimate.

F

Were merged in some form,, so for some of them it might not be the whole thing,.

F

But they seem to me merged in a sense that I found some commits.

F

That sounded vaguely like the project proposal in the time frame, given,, which was as far as I could dig,.

F

And we also seem to have twelve contributors who contributed significantly.

F

After the end of the gsoc, so you can see in some cases that I counted like,.

F

Where was that guy,? There was someone, for example in september, who had some commits, so that's outside of the strict gsoc timeframe.

F

What are the dates on the red column?? The very last commit that is in any of the repositories.

F

So,, I don't know. in german,. We call this "suffering on a very high level.".

F

Seems to me, anyway., I mean,, I'm all up for fixing it,, but now that I made this list,, I don't see the problem.

A

Well, some are like 50-50, right? for merged,, just looking at.

A

Your 12 out of 23?.

F

Well,, do we have any googles here who can say what the retention rate.

F

On google internships, is?.

E

I have no idea what the retention rate is.

E

I think some projects do better and some do worse..

E

I think we're actually not that bad in terms of retaining contributors.

E

Well, in terms of retaining interns,, you wouldn't be taking-- oh,, you mean... yeah,, yeah yeah,, just normal google internships,.

B

Yes., nowhere near 50%, that's what I had hoped for. So we could compare against that.

F

Because it's, for my mindset,, it's basically they're paying internships.

F

It's basically a summer of paid work,, so it's an internship.

F

I did also try to run some statistics,, so this is the time scale normalized to start.

F

On may 1st of the relevant year.

F

The dots are color coded so that the green ones were merged, and the black ones were so-so, and the red ones were failed,, and my conclusion is that there's no conclusion.

F

(laughter from audience), except that people, you can apparently draw the conclusion that.

F

Only projects that were merged had some contribution, beforehand.

F

But this doesn't hold in reverse, so, yeah.

F

I did the same on mail,, though I should say that it's even less precise.

F

For one thing,, I didn't have an egit list copy offline.

F

For another,, it's harder to match on the actual usernames, because people have various addresses and stuff,, but it's similarly,, no conclusion.. So that's for statistics.

F

Most of it has been rehashed.. I made a summary of the thread.

F

so, okay.. We saw this..

F

So these were the theories in the first thread.

F

I think some of them we could probably write down.

F

One I particularly agree with is that we should stay far away.

F

From political projects,, so it's your coinage,.

F

But these are projects where the important part is convincing. Other people.

F

That this change is not breaking others' workflows,.

F

And not actually the writing of code.

I

There's a point of diminishing returns.. My impression from your first flight was that things were getting better in.

A

Recent years than they were,.

F

I think this was my point.. What I meant was that as git matures,.

A

There's less of this frontier, this border, and everything becomes.

A

A little more core,, a little more ossified,, it's harder to change, things.

A

I don't know if that's true or not., I'm glad that's listed on the theory page., because like, shit, I write in an email, like in the middle of the night,.

D

Is not necessarily well thought, out.

F

This was meant to be a summary slide,, so I didn't really.

F

Try to value the theories., I have more summaries., so these were the ideas in the thread,.

F

I think some of them we've already heard.

F

The last one might be interesting. there's this idea that we should.

F

Start again of sorts, and do this in a small scale,.

H

I think that depends on the mergers..

H

Sure., probably also give it all to libgit.

F

You always have a co-mentor,.

H

Which is really good, because staying responsive.

H

Over three months of time, is pretty hard when you have a life,, so you want to take something off,.

H

Who has the time for that,, so I would strongly advise you get a mentor.

H

So, that maybe sets up a limit on the projects, like when we have like three core matches who volunteer and two matches,.

H

That makes two partners.

J

And we took students,, so this was not a limitation by the number of mentors who...

A

But it's true, yeah.. We always have more mentors.. I volunteer to mentor every year and no one ever chooses any of my projects,.

D

Which I think I'm over back on the theories page,, I think the top one.

A

Is me, but yeah,? I think we do have a lot of mentors,.

A

But often the area overlap is not as good,.

A

Like I wouldn't really want to mentor somebody on some modules,.

A

Because I try to ignore them as much as possible,.

A

But that's never...

D

We actually have two good mentors.

A

I would just worry with the co-mentor situation,, it's nice to have a back-up,, but that back-up's only useful if they're, almost as good as the first mentor,. If that makes sense.

H

No,, that's not true. no?, because I was not that proficient with the area,. I tackled it when.

H

I was co-mentoring thomas,, but when he left for holiday, we talked about what's to be done.

H

In what areas, and I had some people like you.

H

So,, the basic thing is: are you responding to the student?.

H

And giving him the next, push him to the right direction, looking after him, that he pushes the patches out,.

H

And you can do that even without being that proficient in the area.

H

K

But you have to have the skills, and knowing what the [inaudible] is involved.

H

And how the community works, and that's one of the proper things to teach,.

H

And any one of us can do that.

H

okay,, so you would argue that we should basically always have a co-mentor,.

A

Even if the person is not in the area.

H

I think you have to be willing to do that.

H

Right., I think you can even co-mentor for stuff you're, not very proficient in, but know something about.

H

I didn't know much and I still don't know much about the git lock infrastructure,.

H

But I was seeing the patches flying by,. I was reading all the discussions.

H

That those guys had,, so I knew what was going on,, what parts of the code were in question, and I always have people I could ask,.

H

If I didn't know the answer, myself,.

H

Let me check, I'll, ask someone else, and come back with the answer. You needed.

H

that's something I think almost every one of us can do.

D

When I work with interns, actually, tends to work better. If I don't.

K

Know the area very well,, but I do know how the process works. Well.

I

Because then, I learn with the intern. The area.

D

If you know their perspective a lot better. yeah, yeah., you then just go, "it's this!", and you don't.

K

Take all the shortcuts, you hand-guide them on,. This is where you go to find,.

K

This is where I go to learn.. Is this a good way to get people to engage with the community?.

E

If they're not getting the sort of technical expertises that they need from their mentor.

E

And the mentor tells them, "don't bother me, go to the list,.

E

Here's the person to cc, that's actually a much better way to mentor,. I think.

E

So,, our problem is that our mentors don't suck enough.

A

(Laughter), you can be our new prototype.

A

I'll, take care of that.

F

I will suggest that we actually go for several co-mentors and for trying to split the image of the mentor, a bit,.

F

Because what you're discussing now is this guidance mentor,.

F

But what we also need is a review mentor.

F

And I think this will have to be a group,, and this group will have to be big enough so that we can basically guarantee to get the review,, even if it's a fairly large patch, in some useful time. Frame,.

F

Maybe 48 hours,, because we don't have the luxury of being peff and having 15 open series and saying, okay,. I will go and work on something else. Instead.

F

But this is the student's job, right?. The student wants to get this review and work on it. Again.

F

And this is probably one of the areas where I qualify as an example for mentors suck.

F

Michael filled in last year, right? to some large extent.

F

Maybe this will also be easier or harder.

F

I don't know, to motivate people to sign up for one job, but not the other.

I

I have a question about the project scope,, because on the one hand, a small project is certainly much more doable,.

I

And if they get it done, and they have extra time, I'm sure if they're interested they'll just work on something that's nearby,.

I

There's never a shortage of work to do,, especially once you've dug into some topic.

I

On the other hand,, I'm curious,, is it hard to get students.

I

On projects that don't sound sexy? that sound like they're, just some--,.

I

-- yes. okay.. I think there is, actually,, because we have sometimes.

A

Less glamorous projects on the list, and they're, the ones that don't get picked.

D

Year after year, and like the students tend to cluster,.

A

So I've,, like I said, I've, never mentored, because no one picks my project.

A

But I have participated in the proposal review every year since 2008, I guess,.

D

And students cluster a lot.

A

We'll get like a hundred applications and there will be, you know,.

A

Most of those, 90 of those were on two or sometimes three projects.

D

And then, the other ones are usually,, there's, usually ten that are just awful,.

D

That you don't even want to, like--.

I

Do they cluster on the undo-able,, huge scope, projects?.

D

Yeah,, I would say so, because they're the most interesting sounding ones, right?.

D

So maybe you should put "laser" on all the uninteresting...

D

G

Well,, I mean there's always the question,. You want to find someone who will be a match for the project.

G

And the community, particularly, so actually I'm curious.

G

Like for the folks here, who were summer of code, students,.

G

I mean obviously,, you have people of personalities.

G

Such that you were a good match, long term for, kind of becoming.

G

An active, long-term participant.

C

Were there particular things that you were looking for.

G

When you were trying to figure out what project was a good fit for you?.

G

Obviously, you were willing to do things that were not just..., you weren't, just looking for lasers.

G

so,, I don't know., I mean frankly,, I'm interested in finding other people like the students that we had,.

G

Who are not just drawn to something flashy, but also are drawn to every aspect of things.

G

Because becoming a long-term participant means that you don't spend all your time doing. The glamorous things.

G

Right. so,, the most extreme form of that would be to say.

D

We should stop having project proposals, and we should just have people proposals,.

D

And then we should tell them what to do?.

D

Is that even viable?.

A

I mean, as far as google's concerned,. We run it how we want.

D

You know,, we take in proposals, but...

D

Yeah,, I mean, it's almost more important to choose a person who's going to be.

A

A good fit for the community than it is to choose a project that we want to get done.

A

Or that we think that person might be a good fit for.

A

Because a person who's a good fit for our project.

D

But doesn't integrate well into the community is going to fail.

D

In one way or another,, you know.

E

One thing we've tried to do: that's,, maybe it's sort of a failed experiment.

E

Is project proposals,? The ideas pages usually have some gaps in them,.

E

And this is a way of picking students that have that kind of planning ability.

E

And have that kind of big picture understanding.

E

That they can fill it in in a realistic way, and they could do this really ambitious. Thing,.

E

And that they don't choose this really ambitious thing.

E

That's unfeasible is like a really good sign.

E

Of course, that has side effects that are not great--.

D

No, I've been either guilty or can be credited with that experiment. Many times.

D

Because I try to do things like, I'll, put up a project proposal.

D

That's like,: we should make large blob support, better, and the solutions could look like a,, b, or c,, or some combination of them.

A

And I keep waiting for the students, like "man,. I totally understand this problem." and like,. This is what we should do, and I'm like, "yes,. I didn't think of that.".

A

That's never going to happen! (laughter) so, yeah,. I do think we've been very,.

A

We've followed that kind of procedure, and I think it's not the right direction to go.

J

Yeah,, so I think the problem is that we should have criteria.

J

Saying what we are going to do and how we are going to teach each project,.

J

And each people,, each student,, and also criteria for how,.

J

Or if we are going to do a gsoc, and how many students we are going to have,.

J

J

And who is going to decide how many students is the..

J

Using which criteria, and so on.

J

F

I'm sure you know about this, pretty much a given.

F

Google gives us an upper bound, and we can still decide to not take some slots, but the--.

A

In fact, we gave away one last year,. I think.

F

Yeah. but we can't really pick the number.

F

We can always say we're overworked,, but we can't really pick the upper bounds.

A

Yeah,, so google comes, after we've had the proposals for a while,.

A

And we started writing them, we're supposed to come up with a number that says "here's how many we'd like to have, and even in our wildest dreams. We would have "n",.

D

And there are "n", that's what we totally have to have,.

A

It would be ridiculous if we don't get them, and that's sort of the lower bound.

D

So it's basically like giving google an upper and lower bound.

A

For what we want, and they come up with some number usually between there,.

A

Unless our lower bound is really unrealistic.

A

but,, usually our lower bound is like two., and they always give us at least two. so...

F

I guess we picked five of them, right?.

F

There were five of them, yeah.. What was that?.

F

2011, yeah.

F

2008 even more, six.

C

I think 2008. Actually it was easier to get slots, because the program was still starting a little bit.

D

And we had participated-- in mentor, suck?

D

yeah, in mentor, suck. (chuckles) so,, I don't know,. I would also like to discuss on the topic of mentor sucking.

F

This scope creep, issue, so, okay,. I won't switch again..

F

I snuck it up on the theory summary because it wasn't actually in the tread,.

F

This tendency of any git topic to devolve into a discussion.

F

Of how to properly solve this, if git were actually a maintainable code, base. (laughter).

F

For example, I mean, this wasn't gsoc,, but it was a spin-off from gsoc,.

F

Lib lock-l got held up the last time. It was held up.

F

On the grounds that, in theory, it should be possible to somehow wire it into the git and log pipeline,, if you could call those pipelines.

F

And I think this is a danger where we actually, as the [inaudible], have to first figure out.

F

If we think the project is feasible in any way, right?.

F

And this, I think,, it doesn't only happen to the g-sock projects.

F

It happens a lot. you're one of the few people who can anticipate what the refactoring thingy will be.

F

And integrate it into front tiers.

F

L

Doesn't anybody with projects regarding the unclear,, let's say, exit, strategy.

L

Imagine you give a certain topic to a student,.

L

And the worst thing that I saw could happen is that he starts working at it,.

L

And okay, doesn't really finish in time,.

L

But then this topic is blocked mhmm.... This is because we have no good idea on how to deal with,.

L

Okay,, we are getting he's not working on it,, but mail is not responding or so,, and this is also a problem giving those projects.

L

To, for example, areas which are important to us, or me,.

L

Because I have the fear that if this fails,.

L

If a student stops working on it,, then I have a lot of work. Getting this topic back to normal.

L

Because it's somehow occupied.

C

Right., because of some legal reasons, it seems we couldn't get integrated.

L

So that we could copy them, but this was not done in time,.

L

So we stuck on this topic for months.

L

It took a lot of time.

A

I don't think that's a problem for gsoc just because at the end of summer,, if things aren't merged, we're kind of happy to write off the student--.

D

We're not like, "don't work on that,, because there's a gsoc student.

A

Whose unmerged work is still sitting out there.? We encourage people to like, wanting.

A

To look at their work and see if it's good, and that's what thomas did, I mean, that was two years after beau's work.

D

That you basically cleaned it up and submitted it,.

D

Which I mean,, ideally it doesn't take that long.

A

But I think that was great!. It was valuable work, and so then he could have come and worked on it in the meantime,.

A

But it turned out the best way to do. It was to pick up his work where it was...

L

Double, yeah, yeah, yeah.

F

I'm not sure it's such a big problem. and also you said the problem.

F

Of unresponsive students,, I mean, I've mentored two of the projects.

F

In the last three years, that didn't really get somewhere.

F

And I can assure you that unresponsiveness wasn't a problem., so beau went to real life and got a chop at that thing. Mackenzie or something,.

F

Immediately after the gsoc basically,, so he disappeared.

C

Thomas is still sticking around,, but in any case, during the gsoc they definitely weren't unresponsive.

D

We have had unresponsive students., the guy in 2009, who failed.

D

I was sort of,, I don't think I was co-mentor officially,.

D

Sort of pseudo co-mentor for him, and he became unresponsive,.

A

And actually he was pretty up front about it..

A

He was like,, "oh., yeah,, sorry, [inaudible], ... cool with that. whatever.".

A

So yeah, those things do happen, and I think it's just part of the process.

A

It sounds like the situation. You were in,, it was good work, but there were legal... things...

A

Which we don't usually run into as far as gsoc...

A

L

...This guy was working for comedy and he was allowed by comedy rules to participate..

A

I think gsoc is easier for that, because there's a lot of signs,. Of course, the sign up front.

A

F

You have to be enrolled in university, and it's always labeled as a full-time job.

F

Across summers,, so it's like you, wouldn't like,, have another job.

E

We're pretty good at being, sort of impolite in disregarding the value.

E

Of the sweat and tears that went into some work.

E

By either abandoning it, or just re-implementing and sort of ignoring it?.

E

Even in the context,, even to regular summer of code, project....

E

(Laughter) so anyway,, I only have one example of this,, it's that when I was doing this with a remote helper,.

E

This worked out really well, in that someone else. Did, david barr did some of the work,.

E

Completely independently, like right before the summer of code,. Technically it.

E

Very much overlapped with what he was doing, and did that work,, like he helped out with that,.

E

And he did some patch review and upstream with that, and other things during the summer.

E

It worked very well., so I think that can be made into a non-problem.

E

For this thing of students being non-responsive,.

E

I've also run into that, in even [inaudible] schools.

E

These sort of test, the boundaries,, and I think that as a mentor,, I would have liked to have.

E

A little more support from the organization in terms of,: these are the guidelines.

E

That all of our students have to follow in order to be allowed to continue,.

E

And that kind of thing.- and I know doing formal things- isn't fun,.

E

But it might be something that we should do.

A

I think I'm not opposed to having more guidelines, or even rules.

A

I think that's a lot of what happened this year was just,. We came to the realization of a lot of these things.

A

Too close to the deadline to actually come up with those rules on time.

A

And that's why we're on it? This year.

A

so,, I'd love to have this conversation over this summer,.

A

On the mailing, list, and hammer out those rules.

A

or maybe before summer, it's almost summer, now.

A

For next year,, so we're not scrambling in january and february.

A

And trying to have this discussion again.

K

Some of these guidelines,, the summer of code, projects,.

K

Say, like you have to, every couple of weeks,, you should send and email saying what you've been up. To, and...

A

Yeah,, I think that's like way too infrequent, actually., I think,, if you're doing git development as your full time job over the summer,.

A

Which of course, summer of code students should be,, then you should be on the mailing list, almost every day.

A

and I actually think we should be getting people to review patches,.

A

Even though I don't trust the review from a gsoc student., but like, many eyes can spot lots of problems.

A

And I think, as long as they're being mentored in their reviews and they're, not the only reviewer,, then I think that's a good thing to be doing.

A

They should be talking out all these other side, issues and stuff, like that.

K

To use the gsoc eyes to see if something's not clear,, or, like if they're out on sundays, [inaudible].

M

Sending these stupid emails, like, what we did, the commentary, was,.

M

That he could have sent the patches to us, and we would take out the other stuff,.

E

Which is, like white space, (chuckles), and whatever.

M

Maybe that should happen on list, though.

A

And other people can...

A

I'm more than happy to tell people their white spaces.

A

A

No, but yeah,, I understand the point..

A

Sometimes these patches need a lot of work to even hit the list. Quality.

A

But you know, people who aren't gsoc students also show up.

A

And have those problems, and...

H

I believe that is for two reasons. one was that maybe students get anxious.

H

To put stuff on the list, at least I was the first time I did it.

H

So, we ended up doing,, he sent it to us,, we look over it,, so you don't make a fool of yourself in public, and I think the mishaps [inaudible] and with time you see the amount of review you do.

H

Before you send the stuff out, degrades and I think at the end of the summer you send--.

A

--, so that's process, and I think it helps.

H

And it helps to get the noise off the list because sometimes it's just beginners' mistakes,.

E

I think it's a part of your responsibility that you don't flood the list.

E

I think flooding the list is good..

E

I think it helps people who are not the mentor to feel that there's a project happening, and to sort of help out when they can.

K

And we seem concerned that the first couple in the series like to use,.

K

Okay, pretend you sent it to me,, and then you can point out the obvious things..

K

Maybe we can end it sooner.

A

H

I think the first series --.

A

-- yeah,, I understand those students are nervous.

A

Except the first series that you mentioned...

H

Then the next patch series comes out and...

H

So it can maybe use that as a rule of thumb.

A

Of course,, none of this applies to libgit2..

A

What do you guys, do,, like for your gsoc students,.

A

You guys are running everything through call request. Line?.

A

G

And we have them in chat as well.

K

Yeah,, that's true.

G

The group is small, and mostly in chat all the time.

G

We have people on chat, call request.

G

We have tons of small things to get started.

K

I love the kinds of [inaudible] by git existing.

G

So much of our work is just catch-up., we're far less likely to propose a research problem, and far more likely to propose a...

G

The goal you're shooting for is extremely well-defined by git already,.

A

I wouldn't say extremely well-defined,, but, yeah.

K

Either it does what git does, with all the work through the wrinkles, or it doesn't.

G

And it's compatible with git, or....

G

That's actually an interesting thing that we haven't talked about, the google summer of code students on libgit2, so far,.

G

Have worked primarily on libgit2 itself.

G

But we actually have a bunch of binding folks here, and it's one way if we wanted to expand the projects.

G

Out of c into other languages,.

G

Plenty of the bindings could use lots of help,.

G

And it might be an opportunity for someone who's, not a c expert,.

G

But who wanted to participate in git to make some really significant contributions.

G

In binding land. I know certainly the bindings could use a lot of help.

E

You mean, in terms of things like iterators? or...,.

G

Yeah, anything in ways of bridging the libgit2 functionality.

G

Into something that will feel natural in the other language in question.

C

You can't get somebody who has no c knowledge,.

G

But somebody who's passion is making r.

G

In a language, right?, I'm sure that the folks wouldn't mind.

G

I don't know if bindings are written in c, still.

G

It can't be not c, but...

K

But it is sort of a british in-between.

G

If it's someone who doesn't feel as comfortable, there's plenty of work where the language--.

I

The language then, with git.

I

the community work, is on the side of the scripting language, right?.

G

That's where it would help, right? it's where the creator's work is understanding. What feels natural in that language,.

G

And what is going to be something that that community is comfortable with,.

G

Which,, I can't tell you what they want in their binding,.

G

But someone can. hopefully.

A

I'm going to put a native perl binding on the...

A

Oh,, that is top priority.

C

We need someone to look out for the amicable...

F

But I think you sort of implicitly point out an important point in this whole discussion,.

F

If we make up a formal, requirement, and then libgit2 has some other workflow, where this formal requirement doesn't make sense,, it will get very hard to enforce the formal requirement.

F

for example,. I would be all in favor of saying something like,.

F

If you fail to report an update in two weeks across the entire summer,.

F

The org.admin (that's, the chief asshole),, will fail your project.

F

Despite what the mentor might say about it.

F

I think they'll be fine,. I think you just have to leave room to say.

A

What form that feedback takes for the gitcore projects, yeah, should be on the mailing list.

A

It doesn't work, it's a different project,.

K

Although I need to send that every month or so,, as in I'm in like an updating, hell.

D

...Three times.

K

More often, it doesn't make sense.

A

So, out of all those projects,, most of those are four git projects.

D

But we've got what? four libgit2 projects, and there was the one egit project.

D

In 2008, I thought there was another jgit project.

D

Didn't one mentor something last year?.

F

Go back to the disclaimer, also a bit., apart from the very first libgit2 thingy,, which was hard to find information on,, it seems that they were all stellar, right?, so statistically speaking,. We should give the slots to libgit2.

F

I'm happy to keep libgit2 under the umbrella of git core,.

A

I've been happy with that,, I mean, we usually get one slot,.

A

And it's been usually good projects and obviously libgit2 does like catch up,.

A

And I like that, the communities are talking,, but.

A

It would also be fine for libgit2 to apply separately to summer of code,.

A

And I think, there's even a form on the application.

A

That's like,: is there some project that can vouch for you.

A

Kind of thing, and if libgit2 wants to apply next year and say,.

A

Yeah, it can vouch for us,, then they probably ask me,.

A

Why is libgit2 being separate? and we can explain to them and I think they probably would give you a slot.

K

We did consider this year, but it didn't work. Out.

A

K

That may be a purchase because it would simplify.

G

The parameters of what acceptable contribution...

A

Right, exactly, it lets you have an org.admin who handles everything..

A

I've been the org.admin and you don't actually do very much or maybe that's part of the problem, that we don't...

A

(laughter). It's mostly about reviewing proposals, and then making sure everybody's good,.

A

Then, of course knowing what they need to do.

A

Students know what they need to do,, and if anybody has problems,, they can come to you with administrative type things., but with this libgit2 stuff,, I just don't pay attention to it.

A

And I just assume that the libgit2 mentors are doing fine or whatever,, so you'd really do the same thing..

A

It would just be more formalized under google's way of looking at it.

B

Just out of curiosity,, what do we get from summer of code?.

A

Oh,, it's like a drain on the... (laughs).

B

The question is,: would we be better off running our own internship program.

B

That is year-round type thing.

A

Oh, yeah,, we could do that.

B

We could have more control over it.

A

It's basically, we get the money,, it's a well known, program,, it's advertised very well, and people know how to find git,.

A

And the mentor summer.

B

That almost seems like that's almost the downside,, where it's like, students are like "I want to make some money. This summer",.

B

Or something,, or my professors told me that I should probably go, do this, and they just go through all the things, rather than being specifically interested in git for some reason.

B

yeah,. We get a lot of applications that are clearly like that.

B

so,. I mean that's another thing that might be, or even like a useful fear. If we did like something through the agency.

B

E

Also, this, in the previous series of the summer of code there was, students could apply through a maximum of 20 organizations,.

E

And this year it was five.

E

which I think really helped.

D

Yeah, I haven't seen, because we're not in it this year,, so I don't know...

D

so,. I would be slightly careful about the lawyer, stuff.

F

I mean, I'm no lawyer, and I'm not an american,, but from all the regulations and all the faq's in gsoc that relates to.

F

Legal issues, check those out first before you just say,.

F

"I will run a program like this.", I mean, it might be feasible if you decide in just--.

I

-- recruiting worldwide, right? if you're, recruiting only in the us or somewhere, where you know...

I

Yeah., it's what I wanted to say.

F

If you stick to us people,, maybe that works.

A

Yeah,, but we definitely would want to recruit from europe, at least.

A

Actually,, I was thinking the opposite,, not the opposite.

B

I was thinking it would be interesting to have the program be open to people in the southern hemisphere., because it's not when they're off school.

E

Oh,, that's true, yeah.

B

So there's basically half the world that can't participate in google summer of code because it's not during the time that they're not at school.. But I mean,.

B

But like for what types of liability?.

B

I don't know, my gut feeling is that it's a huge amount of paperwork.

D

If you looked at the paperwork for gsoc, yeah.,.

G

Because there's compensation involved...

F

Technically, you're employing people from across the world.

A

I think it would be awesome to have a winter of code.

F

(Laughs) winter is coming.

I

Is there a list of proposed projects, that's kind of being gained.

I

Or easy to add things to?.

F

There are several. (laughter).

A

It was on the perl.org wiki for a long time,.

A

And then the perl.org wiki was not available for a very long period of time,, including the application period. Last year.

A

So I moved that stuff over to a wiki on github,.

A

But now the perl.org wiki is back, and thomas says the wiki, that's like--.

D

-- for practical, reasons.

A

So yeah, I think, so yes,. I can point you at it,.

A

But it needs probably a more official spot,.

A

And I think that related to that, is we probably need a more official wiki,.

A

Or it doesn't have to be a wiki per se,, but the perl.org ones,, I don't know,. I don't go to it,.

A

Maybe it's really well maintained,, but it seems like it's not super-well maintained,, but maybe that might not be--.

K

-- a couple of days ago, there was an email on the list.

K

About one of the--.

A

--, I don't know., I don't remember.

F

Seems to be a recurring topic.

E

F

I don't know., I think that wiki thing could be a topic to discuss.

F

Just for itself.

F

yeah,, we'll put it on the board for talking about after lunch.

A

Are we actually ready for lunch?.

F

It's 20 past. ok.. I don't know,. Where do people want to go??

F

Since we have these amazing four ideas: here?.

F

(Quiet chuckles from audience).

E

There's more about that on site, aswell.

E

No no,, not the talking,, continuing this gsoc topic.

F

I think if you want to participate next year,, I think a lot of this,.

D

Somebody needs to take the lead on writing out some guidelines.

D

And recommendations, and we need to do work collaboratively.

A

On the project page.

F

So if you care enough to raise the idea, so?, okay, good.,.

A

I mean I'm just the org.admin,, I don't know,. What do you think?

A

F

And by the way,, if anyone else wants to be the org.admin next year,.

A

The position is open. like every year,, but no one has yet taken me up on it.

F

Finally, I can summarize and hashtag something and then you can flame it.

F