[adding ietf-822 back in as this is getting into their territory]
"Bruce" == Bruce Lilly <blilly(_at_)erols(_dot_)com> writes:
RFC1036 specifies the syntax for the Message-ID header as follows:
message-id-char = <any US-ASCII char in the range 0x21 - 0x7E except 0x3E)
message-id = "<" 1*message-id-char ">"
with the additional constraint that message-ids must match the syntax
of RFC 822 _in addition_ to that of 1036.
Bruce> That's not precisely what it says, but close enough for many
Bruce> purposes.
it certainly implies (both by the comments in that section and by the
general statement at the start of the document) that the 822 syntax
must be followed in addition to the 1036 syntax, and what I wrote
above is just the combination of what 1036 does say about message-ids.
Bruce> There are some subtlr differences, though; 822 rules out
Bruce> specials except for quoted-strings, while 1036 says nothing at
Bruce> all about specials or double-quotes. So apparently
Bruce> <<;:,"..[]<()@foo.com>
Bruce> would be considered valid according to 1036, bur not 822.
correct, though several systems (such as INN) will reject it, and some
systems that would otherwise handle it correctly will reject it for
local posts (only).
what transition plan?
Bruce> If '/' is now to be unrestricted, there should be mention of
Bruce> that as a warning to implementors of the change.
Since that warning was already in 1036 (which is what, 16 years old
now?), this seems superfluous.
Bruce> Ideally, the differences which 1036 stipulated on top of 822
Bruce> ought to be phased out,
arguably they should instead be adopted by the mail standards.
Bruce> Perhaps, if a convincing case can be made.
2822 did specifically make the message-id syntax more restrictive than
822 (822 allowed whitespace in quoted strings, 2822 only allows it in
quoted-pair (i.e. preceeded by \); I do not know if the latter was
intentional on the part of the DRUMS group, I certainly missed it on
several occasions)
Message-ids that are valid under 2822 but not valid under 1036 (or are
longer than the 250-char limit) seem not to occur in mail with any
frequency (at least I don't see any in those mail collections I have
searched). I would be interested in whether anyone with access to large
amounts of real-world email can confirm this.
--
Andrew.