Re: Negotiated noncompliance

[Eliot Lear <[e-mail]>]

I would like to see discussion open as well.  And since I'm number 4 on
the list as far as I can tell (John, Keith, Paul, me), I'm gonna open it
right here.

I agree with the sentiment behind the wording.  May I propose an
alternative, to be inserted in or around section 6.3, as the editor sees
fit?

--
This standard takes a stricter tone than its predecessor.  It does so
with the experience of nearly twenty years of odd interoperability
failures, and with the realization that where-as thousands of people
used SMTP at the time of RFC 821, now millions of people rely on the
standard.  Non-conformance, therefore, can be far more costly.

If one implementation in an SMTP conversation demonstrates its failure
to comply with this standard, the other implementation MAY vary from
this standard if its developer believes it has sufficient knowledge as
to the intent of the other side.  However, it does risk misinterpreting
the intent of the other implementation, which may lead to altered or
lost messages, protocol deadlocks, or other failures.  Thus great care
must be taken in such circumstances.
--

This is not a matter of "negotiated noncompliance" but more a matter of
dealing with an implementation's failure to comply with the words in the
document.

John Gardiner Myers wrote:
> 
> I propose the following additional wording for resolving the "naked LF"
> issue:
> 
> Section 2.3.7, in "Conforming implementations MUST NOT recognize or
> generate any other..." remove "recognize or".
> 
> Add:
> 
> 6.4 Negotiated Noncompliance
> 
> A general protocol implementation principle may be useful to
> implementations which wish to interoperate with particular types of
> noncompliant peers: If an implementation receives protocol which is not
> permitted by the specification and the specification does not mandate
> the response to such illegal protocol, the implementation MAY consider
> the peer to have negotiated a nonstandard protocol.  At that point, the
> implementation is not constrained by the protocol specification in its
> handling of that particular connection.
> 
> For example, if an SMTP server receives commands terminated by the ASCII
> character "LF" but not containing the ASCII character "CR", it MAY
> consider the SMTP client as having negotiated a nonstandard variant of
> the SMTP protocol wherein the <CRLF> sequence is encoded as a single
> "LF" character.  The prohibition against recognizing the sequence
> "<LF>.<LF>" as the end of mail data indication would then not apply to
> that particular connection.
> 
> Note that this negotiation does not release the implementation from the
> conformance requirements in any other protocol session, including any
> session that relays mail received from a negotiated nonstandard protocol
> variant.

--
Eliot Lear
[e-mail]