I had an awesome professor for a QA course at SFU, who for a project had people choose some piece of open source software at their leisure and work on something

I had an awesome professor for a QA course at SFU, who for a project had people choose some piece of open source software at their leisure and work on something substantial for it related to QA. (Adding a different test suite, improving the existing tests, writing some documentation, etc). There was no pressure to actually send the work upstream, but people doing so made him happy. I thought that was a good model, where you get to spend some time with a lived-in software project, you discover the wonderful world of sending patches if you want to, but in general nobody's time is being wasted. This reminds me of that in some ways, and I think its success depends a great deal on how people are incentivized. There's a tremendous amount to learn from doing this - and it is nice feeling like your assignment is actually something useful - but it's really important people don't feel like they have to get some big change merged to get full marks. That can lead to unpleasant situations.