On Thu, Mar 29, 2007 at 09:11:29AM -0400, John C Klensin wrote:
--On Thursday, 29 March, 2007 14:49 +0200 Arnt Gulbrandsen
John C Klensin writes:
RFC2821 specifies, as "recipients buffer" in 126.96.36.199, that a
minimum of 100 RCPT commands must be accepted, and is quite
specific about what happens in limits are exceeded.
I thought that applies to recipients for the same message.
The other case is not covered AFAICT:
That is correct. I had understood the earlier message to be
asking about recipients-per-message, not messages-per-connection.
However, I think the conclusion is the same: since there are no
restrictions on this now and imposing any limits (or
announcement of limits) in the client-server transaction would
require a new extension, it is probably out of scope. I imagine
we could include some general guidance if there were
sufficiently strong consensus behind it, however.
I am not asking for a new extension, just clarification of
the specification so that Joe Random Spamflighter can easier
understand what is specified, and what is not.
Perhaps there should be a new Appendix that tabulates limits
in condensed manner making checking them simpler in style of
IEEE 802.* Conformance Pro Formas ?
It would list limits and references to main text for rationales.
Somewhat in style of RFC 1123 "SMTP REQUIREMENTS SUMMARY" table ?
That table is more about "feature checklist" than protocol
parameter profile listing, but applying the idea.
.. and of course, there is completely missing a text about
how many MAIL--DATA transactions there are expected to be
supported in a given connection, and how long that connection
is expected to be alive. It can be inferred, but explicite
telling is better.
Also, a style issue is that when a chapter is defining more than
one thing, that chapter should have numbered sub-chapters.
It is far easier to point to somebody that "your filter is
violating expectations at 188.8.131.52.5 of rfc-12345" than
".. at chapter 1.2.3 paragraphs 7 through 9 of rfc-12345"
Similarly, lacking the detailed parameter conformance checklist
means that one has to go through the main text, and to spot
there all small details in very lengthy chapters.
Just my opinion.
/Matti Aarnio <mea(_at_)nic(_dot_)funet(_dot_)fi>