The problem with top-down parsing is that it's slow. The more states and rules, the slower it is. Perl's grammar would take ages to process by a top-down parser.

/s

Parse::RecDescent matches well with the way a lot of people think about parsing, which is one of the reasons it's so popular, I think. (That and the fact that Damian writes darned good software) Unfortunately recursive descent parsing is just darned slow in general, so we can't use it for Parrot's parser.

Parrot's parser will likely be more like the classic parsing tools. They're classic because they work and work quickly, and this is one of the places where programmer convenience will get sacrificed for speed.

Okay, I've done the YACC thing before, and while it's not quite as intuitive, it's not so bad. Furthermore, if most of the slowness of RecDescent is from its being top-down, not from its being full-featured and in-perl, then recompiling the parser after each syntax block should be considered possible. (yacc perly.y; gcc -c y.tab.c) takes about 1 second on my laptop, and since this feature will be more useful in major scripts than one-liners, since we can probably cache standard-library transformations, etc., performance problems don't put this sort of thing out of reach.

Of course, I'm not volunteering to write the code here...

/s