Hector Santos wrote:
Are you saying that you have included a statement in the new
text that suggest implementations intentionally use MULTI-LINE
welcomes in order to trap bad guys?
Nope, that was only a minor point in the discussion. The new
ABNF will be obvious, something in the direction of Tony's ABNF,
maybe using "=/" instead of only "/" for the alternative -- the
latter didn't have the "strongly advised" grouping -- and using
a defined <Reply-text> instead of an undefined <text> unrelated
to the 2822 <text>.
The old prose could be interpreted as if it limits multi-line to
command replies, so probably John would now just enumerate known
cases where multi-line is okay explictitly adding the "greeting".
In theory John could add a note that clients are considered as
broken if they don't accept multi-line greeting, but IMO that's
unnecessary, the new ABNF is obvious. It's no convoluted case
of "MUST/SHOULD accept" plus "MUST/SHOULD not generate", where
an explicit explanation in the prose would be required.
Frank