Both parsers that were presented at the hackathon did clever tricks to be able to do context-dependent parsing, by enabling and disabling rules on the fly. Andreas and I discussed building a formalism out of this. Turns out that someone already did this, for his PhD thesis: http://www.complang.org/colm/thurston-phdthesis.pdf

Using something like this could eliminate much of the informal descriptions as suggested at the hackathon.

Cheers, Jan Paul