Re: #1029: USEFOR 3.2.2 References: Should comments be a MUST NOT? (fwd)

[Frank Ellermann <nobody@xxxxxxxxxxxxxxxxx>]

Bruce Lilly wrote:

> To borrow a statement from another list, we also don't have
> a general rule against drinking gasoline.  Sometimes the
> obvious does NOT need to be stated.

In that case it's not so obvious:  We know that some UAs have
difficulties with excessively long References.  We know that
some servers reject excessively long References.  We know that
I might be able to trim References manually, but we're even
less confident about UAs.  We have a simple followup-and-trim
algorithm, but we're far from sure that all UAs get it right
in the presence of comments.  And of course we know this...

> msg-id = [CFWS] "<" id-left "@" id-right ">" [CFWS]

...and so the implementation strategy "grab <msg-id> content
and add it to the old <msg-id-list> separated by CFWS, then
trim it" might introduce [CFWS] found in the <msg-id>.  Stupid
or not, shit happens.  In that case it's not really broken, it
is only stupid, potentially causing harm for the next (old) UA
that didn't expect the legal [CFWS] => the first UA SHOULD NOT.

> BTW, there is no msg-id-list,

In some cases where I disagree with Charles' ABNF I use my own:

   references  = "References:" SP msg-id-list CRLF
   msg-id-list = [CFWS] *( msg-id CFWS ) msg-id [CFWS]

> -list rules usually imply comma separators

A list can be:  item *( delim item )  or:  *( item delim ) item

Among other ideas, let's not start with LISP.  And delim can be
what it takes to separate items, if you're lucky nothing at all,
in fact this _could_ work for a msg-id-list, but we want a CFWS.

Check out <path-list> (no comma), <key-value-list> (semicolon,
in the SPF-1 RFC proposed here to get <injection-info> right),
I won't dig through more RFCs to find lists not using a comma.

> I find it hard to get worked up about a lack of second-hand
> interoperability problems related to comments which are
> rarely inserted in a References field.

Not rarely, it used to be NEVER based on s-o-1036, and we want
to change this adopting the RfC 2822 format. Carefully and not
wholesale, e.g. "we" (TINW) want a CFWS instead of 2822 [CFWS].

> in the grand scheme of things a
> you-don't-have-to-parse-obs-references-because-MUST=MAY
> contortion is going to do a great deal more harm than a
> comment.  Well, that's the law of unintended consequences
> for you.

That's "when Bruce starts to rant there are no more serious
bugs and showstoppers" for me... <gd&r>

                           Bye, Frank