[Top] [All Lists]

Re: Blind acceptance - draft-hutzler-spamops-00.txt

2004-06-15 01:32:01

----- Original Message ----- 
From: "Matthew Elvey" <matthew(_at_)elvey(_dot_)com>
To: "ASRG" <asrg(_at_)ietf(_dot_)org>; <ietf-smtp(_at_)imc(_dot_)org>
Sent: Monday, June 14, 2004 11:58 PM
Subject: Blind acceptance - draft-hutzler-spamops-00.txt

From the I-D "Email Submission Between Independent Networks",
which hopes to be a BCP:
"Similarly, an MDA can choose
   to reject all mail to recipients for which that MDA has no
   arrangement to perform delivery."

Indeed, this is best practice; it should be added to the
"Best practices are:" section, i.e. suggested text:

o  MDAs SHALL NOT accept mail to recipients for which that MDA
   has no arrangement to perform delivery.


Hi Matthew,

At first, as I read the document, I believed this was somewhat implied with
both the first and second BCPs in section 3.0.

However, I believe see the distinction.  Do you mean instead?

o  MDAs SHALL NOT accept mail to recipients for which that MDA
   has no arrangement or a restriction is in place to perform delivery
   with the local receipient or target domain.

I guess it is good so spell it out, but keep in mind, as it is statement
(yours or mine) which implies an enforcement of SMTP level local user
validation (LUV) either at RCPT or the DATA stage. This, in my view, has
been the major source of conflictive debates and opinions which complicated
the "optimal" design of anti-spam implementations.

For a MDA which does not perform LUV at SMTP, a post process LUV is
required, hence contributes to the bouncing issues exploited by SORBIG-based
viruses.   This is something we need to greatly reduced, however, many
systems can't (for one reason or another) can not perform SMTP level LUV.
So probably, the better BCP statement would be:

o MDAs SHOULD consider very strongly performing transaction
   level (SMTP)  local user validation to avoid the high potential of
   delayed bouncing requirements generated by post-smtp local
   user validation concepts.

What we need to keep in mind and also make sure that we "guard" to the
fullest extent is the concept of mail rejection is a traditional and legal
acceptable practice at the transaction level before the message is accepted
for storage.  Message acceptance brings on a delivery or bounce
responsibility that can not be broken.   Since the latter is now a major
source of the malicious abuse of virus spreading mail, it is highly
preferred a transaction level recipient/domain validation should be
implemented by MDAs.

Hector Santos, Santronics Software, Inc.

<Prev in Thread] Current Thread [Next in Thread>