From: Bruce Lilly (blilly@erols.com)
Date: Fri May 16 2003 - 11:15:01 CDT
Charles Lindsey wrote:
> In <3EBFB65B.7040601@Sonietta.blilly.com> Bruce Lilly <blilly@erols.com> writes:
>
>
>>Charles Lindsey wrote:
>
>
>>>I allow "From: Joe Q. Public <joe@public.example>" on a MUST accept
>>>SHOULD NOT generate yet basis, to bring us in line with RFC 2822
>
>
>>That is not consistent with 2822, which states that unquoted dot in
>>a phrase MUST NOT (not merely "SHOULD NOT") be generated [2822 sections
>>3.1 and 4].
>
>
> Yes, but that is the sort of wording we have used at all other places in
> our draft where we have to say "MUST accept, but SHOULD NOT generate yet".
> We need to be consistent.
If we're going to use 2822 as a base, we should be consistent with it
unless there is a very strong reason to deviate. I see no reason to
do so in this case, and a number of reasons to be consistent with 2822.
E.g. there exist common UAs and other message-generating entities; if
2822 says MUST NOT generate, then the only safe course of action is to
not generate the construct under consideration.
And we've discussed this in the past. SHOULD NOT is a recommendation, and
sure enough *somebody* will do exactly what is recommended against, and
yet will be able to claim conformance, since SHOULD NOT is not a prohibition.
The steps in a transition to new syntax are:
0. MUST NOT generate, SHOULD accept
1. MUST NOT generate, MUST accept
2. MAY generate, MUST accept.
Step 0 is optional, although it provides a window where old implementations
can remain conforming while being upgraded to support the new syntax. N.B.
that's a different set of steps from phasing out old syntax.
> However, you seem to accept (and I think everybody else accepts) that our
> draft needs to include the Joe Q. Public phrases at least as a MUST
> accept, which is the main principle to be decided.
I have some doubts about whether that syntax is advisable at all, but that's
not a topic of discussion for this WG. If we're going to attempt compatibility
with the common message format, then we should be consistent with that format's
requirements, so for the purposes of discussion within this WG, then yes, MUST
accept is the way to go because that's what 2822 requires.
For the record, I don't see why we're wasting time with umpteen "Phrase Test"
messages; it's not our issue -- that's up to IETF-822 -- we should simply be
compatible with 2822 (unless our charter is to develop a new, incompatible
format), and 2822 says MUST NOT generate.