[Top] [All Lists]

Re: I-D ACTION:draft-resnick-2822upd-01.txt

2007-04-26 15:56:32

Internet-Drafts(_at_)ietf(_dot_)org wrote:

Hi, where's that discussed, is here okay ?

For the <msg-id> I'd like to see a syntax also working in NetNews,
in news URIs, and in Archived-At links when they use a Message-ID.

A "canonical" version of a Message-ID is specified in

It has the following features:

- once created a Message-ID is "opaque", nobody's supposed to
  modify its id-right, even if that's a FQDN as it should be,
  where modifying say mixed case to lower case is tempting.
- Message-ID generators shouldn't use <quoted-pair>s unless they
- There is no NO-WS-CTL in a Message-ID.

Based on these concepts a "canonical" Message-ID is something like

   msg-id          =  "<" msg-id-core ">"
                      ; maximum length is 250 octets

The maximum length is for NetNews and NNTP, and maybe this limit
is irrelevant for mail.  OTOH mail2news would be forced to reject
or discard any longer Message-ID if they can't guarantee that all
gateways have precisely the same idea how to truncate it.

   msg-id-core     =  id-left "@" id-right

That's handy for other header fields like References in a NetNews
RFC, or in a draft about news URLs, maybe it's unnecesary for mail.

   id-left         =  dot-atom-text / no-fold-quote

The 2822upd draft adds <obs-id-left> as in 2822, for NetNews the
obs-* terms (with two exceptions) aren't used.

   id-right        =  dot-atom-text / no-fold-literal

Same idea, 2822upd needs also <obs-id-right> for its purposes.

   no-fold-quote   =  DQUOTE
                         ( "." *mqtext /
                           *mqtext "." /
                           *mqtext mqspecial *mqtext )

Major difference, 2822upd allows in essence any qtext or quoted-pair.
RFC.ietf-usefor-usefor-12 limits <no-fold-quote> to the three cases
where it's required:

1: "." *mqtext  \/  Something starting or ending with a dot is never
2: *mqtext "."  /\  a <dot-atom-text>, therefore it must be quoted

3: *mqtext mqspecial *mqtext

All <mqspecial> constructs aren't allowed in <dot-atom-text>, same
idea as for the leading or trailing dot issues (1) + (2).

   mqtext          =  atext / "." / mqspecial

One reason forcing a <no-fold-quote> is enough.  Anything else can
either as is <dot-atom-text>, i.e. <atext> or ".", or as in the
<mqspecial> constructs:

   mqspecial       =  "(" / ")" /      ; same as specials except
                      "<" /            ; "\" and DQUOTE quoted
                      "[" / "]" /      ; "." doubled and ">" omitted
                      ":" / ";" /
                      "@" / "," /
                      ".." / "\\" / "\" DQUOTE

<mqspecial> is a collection of characters not covered by <atext>,
plus '\\' and '\"' as the two <quoted-pair>s needed in the left
hand side <id-left> for a <no-fold-quote>.  Last but not least it
also covers adjacent dots, that's not allowed in a <dot-atom-text>.

   no-fold-literal =  "[" *( mdtext / "\[" / "\]" / "\\" ) "]"

Major difference, 2822upd allows in essence any dtext or quoted-pair.
RFC.ietf-usefor-usefor-12 limits <no-fold-literal> to the one case
where it's required:

   mdtext          =  %d33-61 /        ; The rest of the US-ASCII
                      %d63-90 /        ; characters not including
                      %d94-126         ; ">", "[", "]", or "\"

That's not the same as <dtext>, it has obviously no NO-WS-CTL.  The
business about the three '\[', '\]', and '\\' <quoted-pair>s is the
same as in 2822 or 2822upd.  No other <quoted-pair>s are required
in the <no-fold-literal> form of the <id-right>.

For its own purposes RFC.ietf-usefor-usefor-12 decrees that a '>'
must never occur within a <msg-id> or rather <msg-id-core> because
that would break NNTP and other news servers and clients.  It's
fortunately no issue for news URIs, where '>' is anyway not allowed
(and if somebody tries %3E or similar they get what they asked for).

I'd like if the one thing that's essential NetNews, the Message-ID,
has the same syntax everywhere.  We're talking about rare corner
cases like domain literals and quoted strings as local parts in a
Message-ID, <no-fold-literal> and <no-fold-quote>.  As far as these
constructs are anywhere used at all they don't need ugly NO-WS-CTLs
or unnecessary quoted pairs.  I'd also doubt that such constructs
survive a serious 2822 implementation and interoperability report.