Re: Reviewing philosophies and assumptions

[Erik M. van der Poel <erik@xxxxxxxxx>]

> The Japanese use of 2022 for mail is well documented, but in Japanese.
> ...
> Well, I am a bit unhappy about talking on issues
> related to Japan, being no Japanese; maybe the people coming
> from Japan could enlighten us?


Akira Kato has asked me to forward this message:


-------------------------------------------------------------------------------


One of the reference which is available to US people is

	J. Murai and A. Kato: Researches in Network Developement of JUNET
	Proc. of SIGCOMM '87 Workshop, ACM, 1987

More precisely, our Kanji mail exchange code is defined as follows:

1) Use ISO 2022 based encoding in 7bit environment and only G0 is used --
	The announsor of ``ESC 2/0 4/1'' is assumed.

2) Designate appropriate code set into G0 (At the top of text,
   ASCII or JIS-Roman is assumed)--
	ASCII is designated to G0 by ``ESC 2/8 4/2'';
	JIS-Roman is designated to G0 by ``ESC 2/8 4/10'';
	JIS-Kanji is designated to G0 by ``ESC 2/4 4/2'';

3) C0 characters can be inserted when ASCII or JIS-Roman is designated
   to G0.
	Not OK:	Kanji C0 Kanji
	OK:	Kanji ``ESC 2/8 4/2'' C0 ``ESC 2/4 4/2'' Kanji

Notes:
1) is introduced because it requires only 7bit transport channel. Prior
   to the 8bit through tty implementation, MSB's might be truncated
   during the transmission. Especially RFC822 only permit to transfer
   ASCII (0-127) characters.
   Processing of Locking Shifts were avoided because it introduce
   additional states.
2) JIS Kanji can be designated by ``ESC 2/4 2/8 4/2'' in ISO 2022.
   However, we do not use ``ESC 2/4 2/8 4/2'' but ``ESC 2/4 4/2'',
   which is also permitted by ISO 2022.
3) This is special rule to JUNET (and current Japanese academic
   networks). This rule makes the amount of backtrack being limited
   to the length of a line when the text is accessed randomly.
4) The differences between ASCII and JIS-Roman are:
		ASCII		JIS-Roman
	5/12	backslash	YEN-mark
	7/12	vertical bar	vertical bar separated in center
	7/14	tilde		upper score
   and the differences can be negligible for most cases in JUNET.
5) Old JIS-Kanji (1978) is still used in the net with designate
   sequence of ``ESC 2/4 4/0'', which is not recommended.
6) As our encoding requires only 7bit channel, we can exchange emails
   across foreign gateways provided if the very ends can handle the
   Kanji according to the rule above. No patch is required to sendmail.
   MH has been slightly changed to handle Kanji; the patch is available
   via anonymous ftp from, for example, sh.wide.ad.jp [133.4.11.11]:
   ~ftp/JAPAN. Other Kanji handling software is also available in
   ~ftp/JAPAN or ~ftp/japanese on the same host.


-- Akira Kato