Date: 23 Oct 1998 06:03:27 -0000
Message-ID: <19981023060327.16144.qmail@cr.yp.to>
Mail-Followup-To: drums@cs.utk.edu
From: "D. J. Bernstein" <djb@cr.yp.to>
To: drums@cs.utk.edu
Subject: Re: syntax errors
References: <199809242034.NAA25508@mail.proper.com> <006d01bdeb1a$24fc3f30$3a020209@mars.watson.ibm.com> <19981009000105.22092.qmail@cr.yp.to> <3623A37A.9B77528B@cisco.com> <19981013224824.3555.qmail@cr.yp.to> <362D0E9C.1FC55478@cisco.com>

Eliot Lear writes:
> Is it your position, therefore, that if one of the primary the goals of
> the document is to encourage conformance to the standard,

_My_ goal is to help future implementors.

> then separating the grammars into send and accept is a mistake,

Huh? Are you suggesting that the upper limits of safe message creation
are an exact match for the lower limits of safe message parsing?

That is factually incorrect. Interoperability problems exist.

(For example, some message generators will print the year 2000 as 100.
Some message parsers will be confused by that.)

As for ``grammars'': Competent writers don't put religious restrictions
on their writing tools.

> given that it
> forces the standards body to either accept bad behavior or require that
> such behavior be rejected?

Telling implementors the truth doesn't mean ``accepting bad behavior.''

Furthermore, there is a big difference between _allowing_ implementors
to reject something and _requiring_ that they reject it.

---Dan
1000 recipients, 28.8 modem, 10 seconds. http://pobox.com/~djb/qmail/mini.html