Pete Resnick wrote:
no ABNF change, wrt domain-literal/msg-id issues. Frank and Charles
are the only people on the list making comments about the issue at
this point
IIRC Russ and Alexey also prefer better compatibility with NetNews
for the RHS of Message-IDs. If you feel that the NetNews concept of
a "canonical Message-ID" is not important enough for 2822upd there's
another way to solve it:
Just get rid of NO-WS-CTL in the definition of <dtext>. Proposal:
| dtext = %d33-90 / ; Visible US-ASCII characters
| %d94-126 ; excluding "[", "]", or "\"
That also fixes the pollution of the <General-address-literal> in
2821bis, and it's nearer to <IPvFuture> in STD 66. The <IPvFuture>
construct excludes more printable US-ASCII characters, not only [\],
also <>/?#(_at_)(_dot_)
Control characters in domain literals just make no sense, especially
when we're talking about fictitious "General-address-literal" or
"IPvFuture" literals. For real IPv4 and IPv6 literals it's already
settled that they don't contain control characters.
I don't think that you need an <obs-dtext> for this issue. You could
just argue that the interoperability of control characters in domain
literals was questionable using NetNews / mail2news gateways / NNTP
as the evidence.
IIRC my experiments with NO-WS-CTL in domain literals of an SMTP EHLO
also showed that this "feature" isn't interoperable. It doesn't work
with NNTP (by definition), it doesn't work with SMTP (in practice),
so let's please get rid of it.
Frank