PLAN

Things to investigate:

 1. Can we arrange so that Ragel for JavaScript operates on unicode
    strings. This would avoid the conversion to bytes which is slow
    and would make the output code more portable since there is no
    standardised 'ByteArray' object at the moment.

 2. Can we use Ragel to generate code that uses the JavaScript builtin
    objects and their functions rather than processing the input one
    byte at a time. This is important for code size and performance.
    When I write parsers by hand in JavaScript I try to use indexOf
    and match as much as possible so that I avoid looking at every
    character. The more you can offload to the regular expression
    engine the better since more execution occurs in 'native-land'.
    The irregexp engine that V8 uses is JIT compiled on the same engine
    as our JavaScript so getting access to this is going to be the key
    for near-native speed Ragel parsing. My instincts tell me that all
    of the information for composing a parser similar to those I would
    hand-write lies within the grammar definition. Therefore, it should
    be possible to achieve similar code from Ragel. This is a departure
    from the Ragel code generators I've looked at so far.

 3. Also worth checking out:

http://weblogs.mozillazine.org/roc/archives/2010/11/implementing_a.html