From: Bruce Lilly (blilly@erols.com)
Date: Wed Feb 04 2004 - 07:19:02 CST
J. B. Moreno wrote:
> Extra stuff added to a header that isn't authorized by the user and isn't
> required by the protocol -- that sounds like a protocol issue to me,
> partiuclarly when the extra information interferes with the proper operation
> of the header.
You mean like "Re:"?
> If you allow "Re^2: " and "SV:" and "Anwr:" where does it end and become a
> protocol issue?
That's one reason not to put "Re:" in the protocol; it leads to an
endless stream of similar hacks -- "cmsg", "FW", "Fwd", "Auto", etc. in
addition to the ones you cite.
> If the client slips "Your mama wears combat boots:" in at
> the start of every Subject and then hides that from it's user, is that just
> a display issue?
It's a client software issue. You are free to use a different client.
If you like that sort of thing, you are free to use that client (however,
expect to be informed that your taste in client software is not universally
shared). And in case you're wondering, it's not an interoperability issue.
You and others can read, reply to, follow-up to an article regardless of
that client software quirk; injection agents, relaying agents, gateways,
etc. are unaffected.
> What the hell is wrong with saying that the structure of the Subject is:
>
> subject-content = [ back-reference ] pure-subject CRLF
> pure-subject = unstructured
> back-reference = %x52.65.3A.20 ; which is a case-sensitive
> "Re: "
>
> It doesn't match up exactly with mail but so what?
What is primarily wrong is that Subject is and always has been _unstructured_.
Attempting to impose structure on an _unstructured_ field is silly.
As for gratuitous discrepancies, they directly cause interoperability problems;
there are many situations where it is impossible to distinguish a "news"
message from a "mail" message (e.g. IMAP, a message saved as a file, etc.)
so the definitions and semantics must be identical.
Other issues include:
* according to RFC 1036, it's not 100% case-sensitive
* it's ambiguous (one cannot tell whether or not a subject field body
beginning with " Re:" has a "back-reference", since unstructured
can also begin with " Re:"), specifically you have introduced what is
referred to as a reduce/reduce conflict, and that is a very bad thing.
* we are specifically tasked with producing a document which is compatible
with RFC 2822 and MIME; a _structured_ Subject field is incompatible with
both RFC 2822 _and_ MIME.
> We shouldn't be adhering to the mail standard to the deteriment of news.
If you don't want to use the existing widely deployed Internet text message
format and the software which is available to handle it, then develop a
completely incompatible format and watch combined news/mail clients
disappear. But please do so elsewhere, because here we're working on a
protocol document which is compatible with RFC 2822.
> And article grouping/threading by the client should be considered just as
> important as grouping by the server
Servers do not group articles by Subject field content. Threading uses
message-ids and references, not subject.