I think I'm in the process of being persuaded on the "charset" issue.
At least, Neil's belief that it is needed seems to be a lot stronger
(currently) than my belief that it isn't.
What would people (Neil & others) think of augmenting the Content-type
header to include character set info? Right now we have
Content-type: type [; ver-num [; resource-ref] ]
We could easily add the charset information into this header, instead of
a charset header. For example,
Content-type: type [; ver-num [; resource-ref [; charset ] ] ]
or, if you believe (as I do) that charsets are probably more frequently
used than some of these other things:
Content-type: type [/ charset] [; ver-num [; resource-ref] ]
If we went down this route, we could restore "text" as the default
content-type, with something like "USASCII" (let's NOT argue over this
string just yet) as the default charset. A common type of European
message might have
Content-type: text/iso-646
A "typical" troff message might have
Content-type: troff ; null; mm
And a European troff message might have
Content-type: troff/iso-646 ; null; mm
Is there an advantage to this scheme over a separate "Character-Set"
header field? Only in that it preserves the idea that a single header
describes the entire content of the message. (Oh yes, and it saves a
few bytes :-)) Are there any disadvantages?
More to the point, are there people out there whose opposition to a
character set header is stronger than mine, and who would like to take
up the argument against it?