Given that it's realistically going to be Fall 2013 before R-Pi systems are going to be available in sufficient numbers in time for typical school budget planning/acquisition cycles, we might as well just go with 3.x. In addition, there's the ramp-up time needed to train just the teachers with science/math backgrounds, much less the majority of teachers teaching science/math who aren't as strong in those subjects education-wise or interest-wise.
The latter can be a real impediment to effective learning - kids can readily tell when a teacher is just going through the motions with unfamiliar material about which they're not genuinely excited. By the time that latter group will receive training, even Python 2.7 will be approaching end-of-life (for bug fixes and support) and the most-used libraries will have all been ported to 3.x. We don't want to confuse the less technically-inclined teachers with multiple versions, at all costs, or they won't be very helpful in achieving our goals. If kids encounter 2.x legacy code when they're out in the real world, by that time, they should be well-versed enough to be able to accommodate the vagaries between various versions (even within 2.x and 3.x).
The best things in life aren't things ... but, a Pi comes pretty darned close!
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!