Some more about Summer of Code, on the Gentoo side.

First a service information, I’ve taken a day off because yesterday I ended up in the ER after three days of headaches; don’t worry, the (urgent) CAT scan shown everything is okay, and all the blood tests are fine too. I just have to consider this a signal telling me not to stress too much out of me.

Following yesterday’s request about SoC, I started wondering about writing something out of last year’s handling of SoC in Gentoo. The part of me that tries to not be too critic of others told me not to write, but I feel compelled to. I already can tell this is not going to be good PR for Gentoo, but as for once we deserve it, and with “we” I mean myself too, as I didn’t want to mess with Google SoC last year (well, maybe it was better this way as I would have disappeared as a mentor without notice, with what happened).

Up to yesterday, the Gentoo SoC page still shown the information about last year’s SoC. Thanks to Joshua the page is now updated to show the 2008 proposed projects for the students to try applying for, please all the interested parties to look into it. More proposals and more mentors are most likely welcome, and if you’re interested in one of these projects, it might be a good idea to start fleshing out some of the details already, so that when you submit the application you have a valid proposal already.

The old page can be accessed still, even if it’s not linked in the archives (even if the URL is not explicitly referenced, it’s not a secret, anybody can find it in the viewcvs). By the time you read this entry, if you’re not doing so right after I write it, it might be possible that the page was changed, if you can’t understand what I am referring to, check 2007.xml revision 1.1 .

Take a look at the page now. Yes it is correct, there is no information about which students took part in the SoC. There is no list of accepted projects, there is no result information. The page was not updated to reflect the outcomes of the SoC, and not even to reflect the actual projects that formed SoC last year.

I’m sorry, but this has a huge FAIL stamp all over it, in bold red characters. I don’t want this to happen again. The first problem one can see easily is the very limited number of people that the team consisted of last year. The page only refers to Christel and Alec. Luckily this year we have quite more people involved in this task and that hopefully will avoid repeating that huge failure.

But what about the projects themselves? Luckily for us, Google archives the data, so you can find last year’s projects on Gentoo’s SoC 2007 page at Google . Unfortunately, when the time came to review the results of these projects, I was unavailable (remember, I didn’t really come back as a developer till mid to end October 2007, while it wasn’t my fault I still feel sorry for having been unable to help the process). I suppose I would be able to judge whether a project brought good results by having know about its completion after the fact. If I didn’t even heard of it anymore, I would suppose the results were pretty shaky…

• the Collective Maintenance project, I really don’t see any trace of around; failed I’d say;
• BaseGUI I also didn’t find any trace about; failed too; sorry Luis, not wanting to pick on you, but you’re the first one in the order of the list to be also a Gentoo Developer, this detracts point to the idea that Gentoo Developers have more chances to complete something for SoC;
• GNAP cross compile support (and same applies to the other GNAP project, as well as SCIRE); I can’t tell about this, I admit GNAP is way out of my usual league of competence, so I’d like to ask for a status about these three projects to the developers involved; a post on Gentoo Planet would be appreciated;
• archfs: heard nothing about this either;
• equizApp, I’m not sure about this either, I haven’t heard it named in quite a while, but as I remember Betelgeuse discussing again of changes in the recruitment process, I don’t suppose this was successful either.
• Python bindings for Paludis this is probably the only project which I heard of when I came back; the interesting notes are that this is a project that is not, strictly speaking, related to Gentoo (or at least to Gentoo Foundation that is the mentoring organisation) – Paludis is explicitly an external project – and the developer was already a Gentoo developer.

What does this say? Well at a first glance one might actually argue that this plays against my request. The only project that completed successfully was one handled by a Gentoo Developer. But I don’t think the main point to judge by the rate of completed project. As I said, my opinion in SoC is that it is helpful to find new good developers.

I think this was a total failure in that regard. The year before we had even two more applications accepted, and quite a few of those had results that actually can be looked at. In that edition, though, we had a lot more students being developers already (and not even all of them succeeded, anyway). I don’t think the better results are to be referred to the higher dev to newbie ratio, but rather to the higher feasibility of the accepted proposals.

SoC is great, it gives three paid months to a student to work on something they wouldn’t have worked on otherwise. But they are three months. The people has to understand that there will be no more checks after those months, and it comes down to either implement something that can be done and completed, or they have to show intent to continue after those months. That is also an important lesson to learn.

In 2006’s edition, the two Gentoo/FreeBSD-related projects were slightly active, and the students seemed to stay around after a while even if the SoC finished. They didn’t make devs, but I impute that also to the not-much-inspiring environment that the Gentoo/BSD project has become again (and that is also my fault, I want to take back the project as soon as I’m a bit more free than I am now).

As for people from 2007 edition, I don’t remember the name of either of the students (with peper’s and araujo’s exception of course).

The projects, judging by the applications, were quite high-shooting. Proposing to reinvent entirely the maintenance process of Gentoo seems like proposing to fix Italian Parliament in a week (people knowing Italian politics, this is your cue to laugh out loud). One slot was allocated to something that I can’t see interesting Gentoo directly (archfs) and one for an external project (Peper’s python bindings for Paludis).

But how should one judge the intent of the students to stay around after SoC ended? Well I admit that is the tricky part. While I referred to “new blood” I didn’t specifically mean “somebody that has never seen, touched or contributed to Gentoo” bur rather “somebody that hasn’t been an active Gentoo Developer”; people who have been active in the community, even by just reporting bugs, would probably stick around longer than people that never used Gentoo. FFmpeg solves the problem in a slightly different way, as you can see on their SoC Wiki page : they give some admission tasks that the student has to complete before its application is considered for acceptance; if somebody is in only for the money, and counts on disappearing, they’ll likely look for something else and abstain from wasting time and slots.

I don’t think is feasible to get the students to pass an ebuild quiz, for instance, but it might be worth to interview them, or at least to ask them “Since when do you use Gentoo?”. Not like it has to be mandatory to use Gentoo, but it certainly would allow to prioritise people who have at least a clue what they are proposing.

I’m sorry Luis for picking again on you, again, it’s just a matter of who is under hand at the moment. You’re not really an active developer; CIA stats for araujo count less than 400 commits. And they are shared with a FreeBSD developer that goes by the same username. It’s not thus a matter of being already a developer to become more active for the SoC timeframe.

I know this post is really a rant, I know it’s not much constructive, and I’m sorry I can’t be more constructive, for SoC, than volunteering to be a mentor this time and to point out the mistakes I think should be avoided. I can’t come back in time, and I couldn’t really help last year to clean up the stuff after the fact.

I do hope that this year there won’t be the same issues again. And having a stricter selection of students is IMHO a very important thing to do. That and being able to judge the feasibility of a project. So as a suggestion to anybody who wants to apply, I can tell this: work beforehand. Even before submitting the application, try to flesh out details, try to understand what is involved. Don’t be afraid to ask, especially questions about the structure of a software or the method used for maintain it.

Read the blogs, read the documentation and check the commit history of the project. Check who proposed a given project, and ask him more details, I know I will be quite happy to give them to you (and to others maybe, through a blog). [Service request for Joshua and the other people working on SoC: if you’re reading this, please add a contact entry in the table, so that possible students can contact who proposed them if there is not a proper project which they can reference to, like for sandbox).

If you start already to at least investigate what you want to do, it is more likely that we’ll see through your intentions to continue working with the project after the SoC months.

In a different tone, I wish that Jakub, some #gentoo operators and some of the forum moderators could volunteer as SoC admins. If you don’t know what an admin is, it is a person that work for the mentoring organisation that is not going to mentor a student directly, but that is involved in the process of selecting the valid applications. I ask this because you are part of the user-facing interface of Gentoo, and you tend to know who is active (in bugzilla, IRC and forums). And the previous commitment to Gentoo is another important information for choosing wisely the applications to accept.