Clifford Heath wrote:
> Benjohn Barnes wrote:
>> One article there [3] seems to imply that while a packrat parser's 
>> memoization gives linear theoretical time, for almost all real 
>> grammars (and inputs, I presume) it's actually not necessary.
> 
> There's a bit of a chicken/egg problem here.
> 
> Existing computer languages have been created to be easy to parse,
> which means they require minimum lookahead- one or two tokens.
> 
> In part, that's because such languages are easier for humans to
> parse as well, and in part because parser generators couldn't
> produce efficient parsers for them... until now. Now that packrat
> is on the scene, we're free to create more natural looking grammars.

That's an interesting suggestion - that languages are as they are to a 
great extent because of the difficulty of implementing their parsers, so 
existing languages don't need memoized parsing. Certainly a compelling 
point. Do you see newer parsing opening up language design then?

I bumped in to a few suggestions of plugable language syntax while 
reading about PEG, and I definitely like that idea.

Cheers,
	B