Message-ID: <QklaYyi00WBwMEVtJh@andrew.cmu.edu>
Date: Wed,  6 Dec 1995 23:10:38 -0500 (EST)
From: John Gardiner Myers <jgm+@CMU.EDU>
To: drums@cs.utk.edu
Subject: Re: My opinion on VRFY
In-Reply-To: <199512070203.UAA01248@jean-baptiste.internetMCI.ietf.org>
References: <199512070203.UAA01248@jean-baptiste.internetMCI.ietf.org>

If the SMTP server encounters an alias, mailing list, or some other
entry in a database indicating that the asked-about local-part is in
fact specifically known by the server, why shouldn't the server just
then reply 250 for that address?  What value is gained by trying to
follow the delivery path past the alias/list/whatever.

If, on the other hand, all local-parts are being handed off to some
other subsystem, and there is no mechnism to query (or a policy
against querying) that mechanism, the server should reply 252.

My suggestion to Eric for some future version of sendmail was to add
yet another mailer flag, controlling whether VRFY on addresses that
resolve to that mailer should return 250 or 252 (unless any passwd
check, ruleset 5 call, etc. says otherwise) .  In normal situations
"local" should be configured to return 250; externally bound mailers
such as "smtp" and "uucp" should return 252.  In certain situations, I
could imagine that things should be configured differently.

As I see it, the difference between 250 and 252 is whether or not the
server determined that a local-part is likely to be deliverable.
postmaster@local.host should return 250; foo%remote.domain@local.host
should (assuming the server implements the %-hack) either reply 252,
or it should send a VRFY command to a server for remote.domain and
return the result.

-- 
_.John G. Myers		Internet: jgm+@CMU.EDU
			LoseNet:  ...!seismo!ihnp4!wiscvm.wisc.edu!give!up