While working on the core is admirable, I think gsoc would provide an opportunity to port important Python libraries to 3.x. It's important to remember that doing ports helps the core immensely by uncovering 2to3 and py3k bugs.
It's a very noble task to have important python libraries ported to python 3.x. I've played almost whole last year with porting important python libraries to work on different python interpreters (running the same version of python). They differ in tiny details, only a bit. And guess what, it was not only hard, but also very tedious. And now consider student, who looks for joy and is facing python library (say medium, couple tens of k lines of code). With help of 2to3 is getting something that almost works on top of python 3. Except for few small details. This probably means couple weeks spend on debugging obscure failures that end up depending on different string representation of exception or something like that. Assuming he knows python well enough to understand not only major differences (which are handled by 2to3 anyway), but also all minor ones. And those tiny which makes you wonder why unicode subclasses and string subclasses are not exactly behaving how they're defined in a spec.
Suppose student is smart and likes debugging and it's all working. Now the question is, who will maintain the resulting library? Will original team of say twisted maintain it or will it be up to student? Will it just rot in a corner? Who'll maintain buildbots for that?
I think we need to ask first guys who spend their live maintaining libraries instead of just proposing "let's make some poor student port it to py3k", but I might be just wrong, I don't know.