----- Original Message -----
Yeah, the parser generator can be easily implemented using Boost.spirit.qi . So, implementing that would not be much for GSoC. But I have another (a bit complex) idea in this field.
Rather than writing a tool to generate tag based DSEL parsers, I can work on one abstraction level up... A tool to generate formal grammars themselves. Why not a library that would take Two level Grammar (Ex : Van Wijngaarden Grammar ) and build parser for the resulting CFG ...
It's established that two level grammars are more powerful than single level CFGs and that two level grammars are Turing Complete themselves.
I would defer to people with more expertise in language theory than myself, but Turing completeness in a grammar can be a double edged sword: http://en.wikipedia.org/wiki/Van_Wijngaarden_grammar#History