The PRUs and huge GPIO count are the good things about a Beaglebone, IMO. It seems like there's less documentation and pre-built libraries out there for them than for the RPi computers, which is a shame.
RPi wins out on price and community.
Long-term, it seems like the goal of the BeagleBone foundation is to support each board for a number of years in order to allow for text books, curriculum, and other support material to emerge. Their focus seems to be on robotics and headless embedded systems rather than trying to duplicate the "cheap desktop learning computer" objective of the R Pi. It will be interesting to see if they're able to achieve that goal.
I do think the BBB model for open source is the most comprehensive model so far. I'd be very interested to hearing of other open source hardware models. The Open Compute Project hardware licenses were promising, but I haven't seen much action on them since the request for comment in 2014... http://www.opencompute.org/blog/request-for-comment-ocp-hard...