On 25/03/2012 11:02 AM, Simonson, Lucanus J wrote: > > Rather than completed libraries we should be focused on getting > active community members out of the GSOC program. From that > perspective, biting off more than they can chew during the summer > is actually a good thing. If a GSOC student stops at the end of > the summer and we never hear from them again it hardly matters > whether they finished the project or not, because they aren't > going to be maintaining the code. To a certain extent, I think > that our odds of turning a GSOC student into an active boost > community member are higher the more ambitious the scope of the > proposal. I agree with this point, long term retention of talent is a very crucial issue. But we have to face the reality that most if not all Boost projects that could be successfully completed within one GSOC period, have already been done over the last decade - there aren't anymore "simple" Boost projects that can be started from scratch, and where there already exists a major desire for said functionality from within the community, to entice someone to make such a proposal - at the end of the day the GSOC candidates themselves want to be successful in completing the project, but also would want to be able to have some guarantees that they will be "successful", if that makes sense.... Hence the two main remaining avenues for a GSOC candidate's contributions are by building extensions upon already established libraries GGL, GIL etc, or contributing to bug fixes and documentation of existing projects - there is a third avenue which is quite rare, and that is the potential candidate has been working on something long before they entered university or whatever it takes to be GSOC eligible, have a great deal of c++ knowledge/experience, understand Boost structure and design concepts, and have something amazing to offer, rare but not impossible - I believe fusion and its refactor was a good example of this. But again none of these guarantee the individual will become an active long term contributor.