Ton attended the Google Summer of Code Mentor summit last week and wrote this report.
149 orgs, 320-something people attended. It took place at Google's Tech Corners, a new campus in Sunnyvale.
I attended a couple of sessions, but most of the time I spent meeting with all the other open source organisations out here. It was great to talk with developers from VLC (they love our movies, want 3D stereo 4k movie, I asked for better frame stepping and scrubbing) Inkscape (small team, stable sw, they need OSX coder), Appleseed render (mostly Max/Maya users, photorealistic focus), FreeCAD (we should work with them on 1-click export from FreeCad to Eevee), Samba, Libre Office (they work on a Google docs version), KDE, Python (next year 3.0 will be really official!), FFmpg (Blender should upgrade to latest they say), Django, Zulip Chat (interesting hybrid of chat+forum+mail), etc etc.
Blender is well regarded and well known in FOS circles. I met with several Blender users there, even one who supported the "free Blender" campaign in 2002!
In general our own experience with GSoC is quite similar to what other orgs had. Some notes:
- In average it looks like our GSoC projects are too complex. It can also be more simple like "dive in module X, bring it back to spec (+ update API docs) and solve or find the bugs".
- Good students will always find things to do anyway. A simpler project definition can easily lead to bigger projects - students get paid for their time, not the target.
- Many orgs had problems finding good students. Still a lot of noise from fake proposals come in.
- Universities (especially with specialist departments) are not much aware of GSoC, or not aware that students can do specialist work with highly qualified mentors.
- Next time, the moment you get slots we should immediately assign students. You can always swap or release. It's a bit of a game now - the first org who picks a student will get it (students are not allowed to know).
- Efficient spending of mentor time is essential. Don't mentor a student when you could code it all yourself like in a few days.
- One org (forgot who) had an interesting mentoring approach. "We don't take any initiative ourselves - we wait for what the student comes with. If the student doesn't convince us in the first month, we just don't pass him/her.
Feedback sessions with Google's open source office:
- In 2018, a student can submit max 3 proposals. Might give less noise.
- Good on-boarding for free/open source projects is essential. It's not a part of GSoC to have students work on that, but it's a crucial feature for a successful GSoC. Google checks on ways to support orgs with this.
- Tip for the ideas page: intro video(s)! You know, modern times, people don't read :)
- Assigning two or three mentors per student is much appreciated
Our main action point for a successful 2018 GSoC would be to have an active campaign targeted at universities with strong CG education (to attract phd students too). And as second action point to already start recruiting now - students who are currently getting involved should be aware of the GSoC opportunity.