Google Summer of Code 2007 kicks off

This article was originally published in LWN on 2007-04-17.

When Google’s 905 accepted students officially begin coding on May 28th, 137 mentoring open-source projects will be given a serious influx of labor. As in 2005 and 2006, Google will be paying these university students $4500 each to act as pro-tempore developers for the scores of participating open source projects. The projects are given $500 for each student they take on and mentor. Google is spending more than four million dollars on the program this year, and significant contributions are expected for projects ranging from Ubuntu and Mozilla to scientific endeavors.

The Google Summer of Code 2007 reached the second major milestone in its program timeline on April 11th by posting accepted student applications. This follows the announcement of accepted mentoring organizations and the simultaneous commencement of the student application period, the first major milestone, on March 14th.

For more information about the Summer of Code’s motivation, evolution, logistics, finances, and success, the Linux Weekly News turned to Google’s Open Source Program Coordinator Leslie Hawthorn:

LWN: What prompted Google to found the annual Summer of Code in 2005? What were the initial goals of the project, from Google’s perspective?

Hawthorn: We were looking for a way to provide students studying computer science and IT another set of opportunities to work in their field of academic study over the summer. We were also hoping to introduce more students to open source development, as well as to increase the number of open source developers in general. Of course, the great by-product of all of these goals is that more open source code is available to everyone.

LWN: Were there any complaints or controversies from past Summers of Code that prompted logistical changes for the 2007?

Hawthorn: We’ll always be working to improve the program. Payments were an area of difficulty in 2005, and also last year, though much less so. This year we’re introducing a whole new method of payments which we hope will mean that disbursements go more smoothly.

We also heard from many of our students and mentors that there simply wasn’t enough time during the program for students to come up to speed and get their code written, debugged and integrated back into the code base. We’ve reworked the timeline substantially this year and added in two months of “community bonding” to help with the time crunch.

LWN: According to a 2006 article by the Internet News, about thirty percent of students continued to work with their mentoring organization after the completion of their Summer of Code 2005 project. Are you pleased with that retention rate? What could be done or has been done to improve it?

Hawthorn: We’re pleased, but we’d be ecstatic if that percentage were even higher. We’re hoping the community-bonding period will encourage more students to stick around after the program ends.

LWN: Google was only able to accept less than half of the mentoring organizations which applied this year. What criteria did you use to differentiate and select these organizations?

Hawthorn: We based our decisions on several factors, but the major area was an organization’s Ideas list: Did it look like a set of reasonable goals for students? Did the projects look interesting? Did the organization’s Ideas list look well-organized and clearly written?

LWN: How does Google benefit from accepting mentoring organizations such as the Robert H. Lurie Comprehensive Cancer Center of Northwestern University, which seemingly have little or no relevance to Google’s business?

Hawthorn: We take orgs from a wide variety of technology spaces. We benefit in the same way the rest of the world benefits: by having more code produced for everyone to use.

LWN: Both mentors and students are required to submit evaluations which Google uses to determine whether or not the participants successfully completed their project and deserve payment. What questions are on these evaluation forms? What factors does Google use to determine whether or not payment is deserved?

Hawthorn: The questions are newly determined each year. While Google makes the final determination on whether a student receives payment or not, in almost all cases we go with the mentor’s judgment call. (We have not determined the questions yet for this year.)

LWN: Can you estimate the success rate in past Summers of code based on these evaluation forms?

Hawthorn: We had 89% of students pass their final evaluations for the 2005 program, and 82% for the 2006 program.

LWN: What is Google’s overall budget for the Summer of Code? How has this budget increased since 2005?

Hawthorn: Our stipends budget has increased $1M each year; in 2005, we awarded $2M and in 2006 we awarded $3M. This year, our funding will increase to $4M. Our overall program costs behind the scenes will increase somewhat year by year, but not notably so.

LWN: Google pays $5000 per student project - $4500 to the student and $500 to the mentor. What other major expenses are involved in the Summer of Code, be they during planning, administration, or otherwise?

Hawthorn: Shipping internationally is incredibly expensive. We spend a decent amount of time planning for the program when it’s not in session as well.

LWN: What do you find usually motivates students to participate in the Summer of Code? Surely it’s not just the money.

Hawthorn: It’s the t-shirt. Seriously, though, students are drawn to the program for a variety of reasons: passion for open source, desire for recognition and potential career advancement. That said, though, you would be shocked at the number of student and mentor applications that specifically mention the t-shirt.

LWN: Is it too early to announce the Google Summer of Code 2008?

Hawthorn: We have nothing to announce now. Stay tuned! :)

This interview was conducted on March 28th, well before the selection of student projects. Following the selection process, it was revealed that Google had accepted several hundred more applicants that had been previously expected. Hawthorn was contacted again on April 15th to reflect on this decision:

LWN: A posting on the Summer of Code FAQ seems to indicate that Google was expecting to accept around 600 students this year, a number similar to 2006. Assuming the budget of $4 million you mentioned earlier was not increased, how were you able to accept nearly 900 students in 2007?

Hawthorn: We asked for additional funding. Many organizations let us know that they had received extremely high quality proposals, so we secured a higher budget so we could turn fewer of these brilliant students away.

What’s coming in the 2007 Summer of Code

While the third annual Summer of Code will appear largely familiar to 2006 participants, the administrative improvements Hawthorn mentioned are significant. The delays she referenced which some participants experienced in the issuing of their payment from Google were among the many technical issues that created friction in past Summers of Code.

The “community bonding” period Hawthorn spoke of began with the announcement of accepted students on April 11th and will last 47 days, until coding begins on May 28th. During this time, students are encouraged to follow the mailing lists, IRC channels, and forums pertinent to their project and open communication with their mentors. Last year, students were expected to begin working on their projects as soon as they were accepted to the program.

The “Ideas list” which Hawthorn suggested was used as a criterion in the evaluation of mentoring organizations is a list of opportunities for student projects. For example, see KDE’s extensive 2007 ideas list. Though some students submit their own concepts in project applications, many choose to support open source projects in the areas which they, themselves, have identified as needing attention.

Oh, and - yes - both students and mentors alike are very happy to receive their free t-shirts.

The number of participating mentoring organizations has grown dramatically since the inception of the Summer of Code. For 2007, 137 projects have been listed, 102 projects were listed as participants in 2006 and just 41 were listed in 2005. This enormous increase in participation comes in spite of the reality that Google had to deny nearly 200 other open-source projects this year.

Among the many mentoring organizations new to the Summer of Code are the the Git revision control system, GnuCash financial application, Haiku operating system, MacPorts OSX application porting project, Maemo Nokia Internet Tablets development platform, MySQL database, the Neuros Technologies multimedia device firmware, OpenICC color management initiative, Plan 9 operating system, Rockbox multimedia device firmware, Scribus layout application, the Electronic Frontier Foundation, the Space Telescope Science Institute, and the WordPress blogging software.

Unfortunately, many of last year’s participants are no longer mentoring Summer of Code projects, including the CUPS printing layer, Daisy CMS, GIMP image editor, the Internet Archive, Mars Space Flight Facility, MythTV, One Laptop Per Child, openSUSE, React OS, and XenSource. Hawthorn notes that the loss of these projects, some of which were among the most active participants in past years, is attributable to both applications being rejected and projects failing to reapply.

While nearly all of these organizations were granted multiple student programmers, some are employing significant percentages of the entire Summer of Code population. After receiving a stagerring 213 applications, KDE has been allocated 40 students this year, even more than their total of 24 in 2006. KDE’s total is the highest amongst participating organizations this year. Other organizations with more than twenty students accepted are, in descending order, Python, GNOME, FreeBSD, Apache, Mono, Eclipse, Drupal, and Ubuntu. Some Summer newcomers amassed impressive participation, including Plan 9 with 13 students and MySQL with 10,

The Summer of Code has popularized the seasonal student internship format among Free software projects following its introduction in 2005. Already, many copycats have begun and concluded with varying success. In cooperation with Google, Gnome sponsored the Women’s Summer Outreach Program 2006 in which six female students were given $3,000 to “hack on Gnome-related projects” for two months.

KDE also operated a Summer of Code-analog in 2006, the Season of KDE, in which sixteen students who applied but were not accepted to the Summer of Code were paired with a mentor and offered the opportunity to make unpaid contributions to the KDE project. Unfortunately, nearly five months after the Season of KDE’s conclusion, nearly all of the participating projects have failed to report completion, with the notable exception of Ivan Cukic’s Kamion user-state migration tool.

Most recently, OpenUsability’s Season of Usability 2006-2007 rewarded six international students with $700 each for reviewing and improving the usability and interface/interaction design of GeeXboX, Okular, and other open source projects.