Robin Houston is a very, very bad man. I've just got out of his Mutagenic Modules mind fsck and he's doing some very scary things. Most of the stuff he presented was proof of concept/scary rather than actively useful, but the possibilities implicit in some of his examples are phenomenal.
And when he got onto the possibilities that open up once he's got all the bugs ironed out of B::Deparse things got really scary.
You see, once you can accurately deparse all possible optrees, you can write a B:: backend that will generate an abstract syntax tree. That you can manipulate. Refactoring browsers ahoy! Perl->Python XSLT stylesheets! Bad obfuscatory scariness! It's all possible, the world is the mollusc of your choice.
And the odds are good that he's going to get it working, because Hugo van der Sanden was in the room and sounded very interested in working out how to remove the last hurdle in B::Deparse's way...
