🚀 go-pugleaf

atkinson@itd.nrl.navy.mil (Randall Atkinson) writes:

>  Shouldn't really take any longer if done correctly.  Certainly the
>good folks at AT&T Bell Labs have shown that it works quite nicely.
>The Bell Labs encoding for ISO-10646 appears to be the leading
>contender for adoption by MIME as the conventional way to encode
>ISO-10646 in MIME email -- and it would be just fine with either 7-bit
>or 8-bit transport (if I understand Andrew Hume correctly).

The Plan 9 UTF format is clearly an 8-bit format. So we would get the following
transformation pipeline:


               1. The Bitmaps of the characters encoded in a local charset
                  with local codes (e.g. Latin 1, IBM CP850, ...)

                              |
                              V       ISO 10646

               2. The numbers in the rage 0 to 65535

                              |
                              V       one of the proposed UTFs (e.g. from P9)

               3. A stream of bytes (8-bit!!!)

                              |
        only for mail ->      V       an 8-bit transparent 7-bit encoding

               4. A printable stream suitable for historical
                  mail systems.

The receiving system has to perform the reverse steps. (Yes, I know
that I used the wrong terminology and that there are ISO standards defining
a much more detailed reference model of the character set universe,
but a poor CS student can't even pay the ISO documents that define
them, so how could I read them ... :-( )

The only STANDARD which we will have in a few months is a mapping
between the characters (bitmaps) to a space of numbers in a huge table.
Unfortunately, ISO 10646 doesn't tell us a lot about how we should deal
with these numbers. We still have to define the second and third step
in the above pipeline. I heard roumors, that the Plan 9 UTF version will
be added as UTF-2 to an ISO 10646 annex, can anyone confirm this? Perhaps
we really should wait as suggested by Randall Atkinson for the final
version of ISO 10646 before starting to include anything new (e.g.
a 16-bit encoding) in MIME.

Will there be other ISO standards (e.g. POSIX, etc.) that define
precisely how to handle the numbers defined by ISO 10646? The Plan 9
stuff seems to be very promising.

>Moreover,
>the transported character set and the displayed character set need not
>be the same.

Of course. But USENET users will have difficulties in understanding this,
as there isn't currently anything like a presentation layer (as defined in the
OSI reference model) that performs the conversion between the local
representation (e.g. I prefer Latin 1 files) and the encoding used on
the network (e.g. any 8-bit encoding on news and 7-bit encoding on
email). This won't cause any news/email interworking troubles, as
every text has to be converted anyway. (Perhaps in the far future of
ISO 10646 terminals and editors, the presentation layer (e.g. realised
by software converting MIME to/from a local format) might disappear again,
as then the local format and the network format will be the same like
this is today the case with US-ASCII). Yes, MIME is a presentation
layer protocol for USENET and SMPT.

Such a model should perhaps more clearly be explained in future MIME 
extending standards.

Markus

-- 
Markus Kuhn, Computer Science student -=-=- University of Erlangen, Germany
Internet: mskuhn@immd4.informatik.uni-erlangen.de  |  X.500 entry available
----- Anyone participating in the use of MS-DOS, Heroin or Cocaine is -----
---- simply not getting the most out of life possible. (Brian Downing) ----