Message-Id: <199602292007.PAA18769@jekyll.piermont.com>
To: djb@koobera.math.uic.edu (D. J. Bernstein)
cc: drums@cs.utk.edu
Subject: Re: comments from a newcomer 
In-reply-to: Your message of "29 Feb 1996 19:54:33 GMT."
             <19960229195433.20101.qmail@koobera.math.uic.edu> 
Reply-To: perry@piermont.com
Date: Thu, 29 Feb 1996 15:07:34 -0500
From: "Perry E. Metzger" <perry@piermont.com>


D. J. Bernstein writes:
> > Requiring SMTPs to suddenly become 8-bit was felt to be unacceptably
> > harmful to the installed base.
> 
> But the current 7-bit requirement is so harmful that thousands, perhaps
> millions, of people are forced to _ignore it completely_.

Sendmail never ignored it and has been the most popular MTA -- I don't
recall it passing eight bit data until the standards recently allowed
it via extensions.

> Why are you deliberately making de jure standards that don't match the
> de facto standards?

None of this is new. The old RFCs all said that mail was 7 bit and
only 7 bit. In fact, they required the high bit if present be STRIPPED.

> > Relaying an 8-bit message was known to 
> > cause failures or other undesirable behavior in MTAs which were 
> > perfectly compliant with the spec in only accepting 7-bit messages.
> 
> I've heard this rumor too, but can you find any hosts now that choke
> on 8-bit messages?

Yes.

> If I can crash a mailer by sending a few cleverly chosen bytes across a
> TCP connection, that mailer is a security disaster.

You can't crash them, but they won't pass the data unmangled.

> > Let us recall that it was not that long ago that many of us used
> > machines which did not have words that were multiples of 8 bits.
> 
> Right, but that's no reason for artificial dissonance between SMTP and
> the underlying transport protocol.
> 
> If you're running SMTP over a 36-bit protocol, for example, you should
> be allowed to transmit 36-bit data. Anything else is poor engineering.

SMTP was not sent over a 36 bit protocol. It was sent over a protocol
that encoded ASCII as 7 bits in eight bit octets. It was, however,
transporting ASCII messages between computers that sometimes could,
and sometimes couldn't, handle 8 bits instead of 7. It was thus
designed as a protocol for sending 7 bit ASCII, not for sending 8 bit
binary.

Perry