Pete Resnick wrote:
Define the left hand side of the "@" in msg-id to be
(dot-atom-text / ob-id-left)
Good
Leave the right hand side of msg-id as it is
(dot-atom-text / no-fold-literal / obs-id-right)
With a new no-fold-literal adopted from USEFOR.
RECOMMEND (or should it be REQUIRE?) that the right
hand side be a domain identifier (either domain name
or domain literal)
I'd like REQUIRE with an additional "SHOULD be a name"
reflecting the RFC 822 rule:
| THE USE OF DOMAIN-LITERALS IS STRONGLY DISCOURAGED
I'd prefer to leave the ABNF of domain-literal the way
it is and limit in the text.
Not good enough for copy and paste into other drafts.
The "canonical" no-fold-literal syntax is quite simple:
| no-fold-literal = "[" *( mdtext / "\[" / "\]" / "\\" ) "]"
| mdtext = %d33-61 / ; The rest of the US-ASCII
| %d63-90 / ; characters not including
| %d94-126 ; ">", "[", "]", or "\"
In essence that says "don't use quoted-pair unless you
must (and that's never the case for IPv4 or IPv6), and
stay away from '>', WSP, or NO-WS-CTL".
I'd prefer it if you also remove '<' from the picture.
The '>' was only the minimal difference from RFC 2822
for USEFOR, nobody needs angle brackets within a domain
literal. They'd require percent-encoding in news URIs.
I'd also be inclined to *not* talk about case-sensitivity
and comparisons since nowhere is it discussed in 2822.
For consistency you could copy the relevant text from
USEFOR, stupid UAs trying to "beautify" the right hand
side could screw up threading by references.
Frank