Re: Just-Send-8 to ESMTP/MIME Transition

[John C Klensin <KLENSIN@xxxxxxxxxxxxxxx>]

>John, I think using x-unknown would be a step backwards.  Basically,
>I would like to have a token that means "I converted this message
>into mime, and I didn't know what the character set was".  I think
>all parties agree about this.

Neil, I want to stress that I don't feel very strongly about this. 
Also, I don't know whether you have read everything between my note and
the present, but Craig has possibly argued the case better than I can
(certainly better than I have).

As I see it, the root of the argument for x-unknown is that, to make any
sense of it, the receiver has to know the sender and what the sender
does.   If the receiver does (which will often be the case), then such
"unknown" character sets are perfectly useable.  If the receiver
doesn't, then the use of such character sets falls somewhere on the
spectrum between "heuristics/guesswork" and "completely meaningless". 
This "know your sender, interpret accordingly" property is exactly the
nature of what we have described elsewhere with terms like "bilateral
private agreement".

Conversely, "unknown", with or without the "x-", cannot possibly be
"well known and well defined".  "Well known and well defined" implies to
me that the receiver, without information outside the scope of the file
and the RFCs and registrations it [implicitly] references, should be
able to precisely and exactly interpret the character set.  I suggested
the "x-" for that reason and because I didn't think "unknown", as it
appeared in Greg's draft (or in a different location in the earlier "8
bit clean" draft) met the standard for registering something that is
laid down in RFC-MIME.

But it seems to me that your other comment:

>in theory, anyone else could
>use x-unknown to be anything that they wanted, and they would not
>be doing anything against the spec

may be more to the point here, and may be a strong argument for leaving
the "x-" on.  On the other hand, operationally there probably isn't much 
difference between "converted message, presumed to be text but with
unknown character set into MIME and labelled it 'unknown'" and
"identified something as 'unknown' about which I didn't have a clue"
(again, with or without the "x-").  If the history is important, it is
mostly important to people, and the right place for that is in the trace
("Received:") fields.   I think.

    john