[Top] [All Lists]

Re: How to handle a lot of character set Content-types

1991-05-03 07:14:00
I think that one of Bob Smart's persistent suggestions is a dangerous
one, but he keeps including it in lists of things-to-do.  Most recently:

Excerpts from transarc.system.ietf-822: 3-May-91 Re: How to handle a lot
of .. Bob(_at_)mel(_dot_)dit(_dot_)csiro(_dot_)au (2590)

(b) add Content headers on behalf of dumb UAs;

Shades of Sendmail, Batman!

Let's keep the MTAs out of the business of ``adding'' these headers. 
That's a UA's job, as now configured.  Let's not try to ``correct'' an
old UA's behavior at the MTA level.  Rather, let's be sure that messages
without such Content headers are interpreted in an old-style fashion,
conformant with other systems and (one hopes) with the RFCs.

You can be pretty sure that your attempts to ``correct'' your dumb-UA's
misbehavior will do the wrong thing part of the time.  As for those UAs
that use good old MAILASCII (as in 822), neither RFC-XXXX nor system
practice should break them.  As for those UAs that use something beyond
MAILASCII, they're new enough that there is hope of converting them to
say what they mean.

For that matter, the rest of Bob's paragraph is pretty difficult to
swallow, also:

(c) Convert transport encoding to 7 bit where necessary and appropriate;
(d) Convert 7 bit encodings to 8 bit when sending to nearby UAs which 
the MTA knows can understand 8 bit but not Content headers. 

Yow!  More MTA majik!  We've been arguing back and forth about (c), but
(d) is downright nightmarish.