I am enthuistic about the real-headers for extended character
set support in message headers. I have seen no objections to the
scheme, and it appears to be workable. Please include it in the next
draft of the RFCXXXX. BTW, when can we expect a new version?
Not true. I object to it. The basis for my objection is simple: I think
using mnemonic in the original headers is simpler, more consistent, and
leads to something that is a lot more usable on an unextended UA.
Implementation of code to handle Keld's mnemonic mapping (using nothing but
his RFC -- I had his code in hand but I did not look at it) took me about
a day. I don't think there is a significant implementation problem here.
Let's look a little more closely a Nathaniel's proposal here. First there are
three things we have to consider --
(1) Initial composition of headers.
(2) Use of headers during transit and delivery.
(3) Reading/displaying headers once received.
Taking these in reverse order, I make the claim that the Real- proposal
is not dramatically different from the proposal to use mnemonic when displaying
the headers once the message is received. Actually, mnemonic simply requires
a way to map from mnemonic to whatever character set is used locally. Use
and display of Real- headers requires a decoder general character set
converter. I think these are about the same in complexity (thanks to Keld's
enumeration). If anything wins, it is mnemonic, not Real-, in my opinion.
(2) is something of a non-issue. Since both schemes only affect parts of
headers that are not involved in delivery and routing decisions, I don't
see a problem with either one. Neither one requires the use of 8-bit
characters or anything like that.
(1) is a problem, however. In order to use Nathaniel's scheme, one must
encode the text used and place it on the Real- header lines. Now what, pray
tell, goes on the regular lines? You have to generate something. Frankly,
I think the best you can do is use mnemonic encoding!!! And if you do so,
simply noting that you have done so means that the Real- headers are
completely superfluous.
Nathaniel's scheme also does not address what should be done about comments
embedded in _any_ structured header. Are we going to have Real- versions
of everything to deal with this? What about comments on a Real- header
itself -- how do we handle this recursive problem?
In contrast, I have seen no cogent objections to mnemonic posted to the
list. Maybe I missed them.
Mind you, if someone comes up with something I'll be happy to back down on
this one. I just don't see the point of additional header fru-fru when it is
not needed and appears to me to have substantial operational problems that
mnemonic does not have.
Ned