Perhaps you might also want to look at my new parser generator, Marpa. Marpa is based on recent improvements to Earley's algorithm and is a general parser generator. It will parse from any grammar that you can write as BNF, without restriction. It handles ambiguity, left-recursion, right-recursion, you name it.
Speeds seem quite acceptable in practical instances. I've written an HTML parser based on Marpa, Marpa::HTML, which is also on CPAN. That distribution contains an HTML pretty-printing utility, which I find useful in my own work.
General BNF parsing has been neglected over the recent past and, despite signs it can be quite practical, has not resulted in any tools in widespread use. I hope Marpa will make BNF useable in the same way that regular expressions are now.