[mew-int 01596] Re: windows 1252

[Kenichi Handa]

I'm sorry for the late response on this thread.

I at first want to clarify these things:

(1) windows-1252

This is actually not a charset but a coding system in
Emacs.  When Emacs reads a file by this coding system, it
decode each byte into one of these character sets:
	ascii, latin-iso8859-1, mule-unicode-0100-24ff

(2) ctext (alias of compound-text)

On conversion, it works not fully compatible with the
specification of X Compound Text because it encodes any
Emacs characters while using an designation sequence for
private character sets (please note that all Emacs charasets
have a iso-final-char).  So, Big5 characters are preceded by
ESC $ ( 0 or 1, mule-unicode-0100-24ff characters are
preceded by ESC - 1.

(3) ctext-with-extensions (alias of compound-text-with-extensions)

It can handle several kinds of "extended segment".  On
decoding, it handles ESC % / N M L ... ^b for what listed in
ctext-non-standard-encoding-alist, and ESC % G ...ESC % @
for UTF-8.  On encoding, it does two-path encoding; at first
by `compound-text', then re-encode what are encoded by a
designation sequence listed in
ctext-non-standard-designations-alist using the "extended
segment".  Currently only ESC $ ( 0 and ESC $ ( 1 are
listed.  Thus only Big5 are encoded using the "extended
segment".


As to the Mew case, I think the following is good.

When it runs under the current Emacs, keep using ctext but
add a coding tag to the file.  Emacs should be able to
encode/decode all Emacs characters.

When it runs under emacs-unicode version, on writing the
file, if all the characters can be encoded by ctext, keep
using it.  If not (because, in emacs-unicode, some character
doesn't belong to any charset that has iso-final-char), use
utf-8.  And in both cases, add a coding tag.  On reading,
check the coding tag at first.  If no coding tag, read by
ctext, otherwise, read by the coding system specified in the
tag.

By the way,

> The one-and-only coding-system which, I found, meets the requirements
> above is 'ctext.

I think iso-latin-1-with-esc also meets your requirements.

---
Ken'ichi HANDA
handa at example.com