Just catching up on some mail. My but we've all been busy...
I wanted to note that you have things just right. RFC733, and it's
successor RFC822, operated under the strict requirement to keep
grammar as simple as possible.
In this case, the requirement dictated a context-independent
lexical analyzer. The result is text that is far uglier than I would
have liked, but...
From: rayan(_at_)cs(_dot_)toronto(_dot_)edu (Rayan Zachariassen)
Subject: Re: folklore etc.
MRC(_at_)panda(_dot_)com (Mark Crispin) writes:
>... how can you possibly deal with
> From: John T. Doe <jdoe(_at_)foo(_dot_)bar(_dot_)com>
>which violates a similiar rule (which I also consider silly) in RFC-822?
I agree it looks like a silly rule, but it does actually have a good reason
behind it (though only Dave Crocker knows if it was the original reason):
if you allow unquoted periods, you've just changed the address syntax
complexity from LALR to LR(n) (i.e. you need multi-symbol lookahead capabil
ity
to parse properly). It makes a big practical difference in implementation
complexity, especially if you want to give sensible error messages.
I will now make one soapbox comment: Many people gave in on many
points about MIME, in order to get it out the door. We need to be very careful,
in these later discussions, to distinguish true broken-ness from
attempts to modify the spec to suit previously-lost battles or other
issues of preference.
(My own hobby horse (sorry Russ) is that I would have wished that
parameters didn't need to be quoted when they have multiple
values, as in foo=1,2,3; but I can't imagine arguing about this
except on the weak basis of aesthetics.)
Dave