Date: 19 Mar 1996 21:59:19 GMT
Message-ID: <19960319215919.26432.qmail@koobera.math.uic.edu>
From: djb@koobera.math.uic.edu (D. J. Bernstein)
To: drums@cs.utk.edu
Subject: Re: lexical analyzer vs. parser vs. ABNF

> Given the pressure to fix the handling of dot,

Could you please explain this in more detail?

I see pressure to handle dots in phrases. There is a massive installed
base of software that will happily produce Joe Q. Public <joe@af.mil>.
There are also some users who type such things without quotes.

We can extend 822 to allow dots in phrases. The benefits are clear.

The only potential loss is implementation complexity. However, I've
exhibited a _finite-state_ analyzer for address lists, taking a reversed
sequence of tokens and producing a list of reversed addresses. That
analyzer is very simple, with or without dots, and I expect it to be
used in all future 822 parsers.

So the long-term complexity disadvantage is zero, and the short-term
disadvantage is heavily limited even if there are some parsers that have
to be rewritten. (Are there any?)

---Dan