Using iso-8859-15 with Emacs21 & Gnus

Using iso-8859-15 with Emacs21 & Gnus

[Raymond Scholz]

Hi there!

I'm trying to migrate to iso-8859-15 until January 2002 (time of the €
becoming legal tender in some European countries).

That's what I've done so far:

* XFree 4.1.0 with iso-8859-15 fonts.
* Emacs 21.1
* Oort Gnus
* (set-language-environment "Latin-9")

My own messages are properly encoded in iso-8859-15.  But quoting
parts of other messages that are encoded iso-8859-1 causes Gnus to
create a multipart/mixed message with the quoted parts encoded in
iso-8859-1 and added parts in iso-8859-15.

I know that it's difficult or even impossible for Gnus to determine
whether iso-8859-1 would suffice to encode the whole message.  What I
would prefer is telling Gnus to encode the whole message in
iso-8859-15.  How?

Cheers, Ray
-- 
Hfr fgebat rapelcgvba!

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Raymond Scholz <ray-2001@zonix.de> writes:

> I know that it's difficult or even impossible for Gnus to determine
> whether iso-8859-1 would suffice to encode the whole message.  What I
> would prefer is telling Gnus to encode the whole message in
> iso-8859-15.  How?

Dave Love has posted a Lisp package for Latin unification to
gnu.emacs.sources.  This might solve this kind of problem.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Raymond Scholz]

I forgot to mention that I didn't use Mule-UCS before.  With Mule-UCS
0.84 utf-8 messages are being created.

Florian Weimer <fw@deneb.enyo.de> writes:

> Dave Love has posted a Lisp package for Latin unification to
> gnu.emacs.sources.  This might solve this kind of problem.

Tried that but it makes no difference.  With or without Mule-UCS.  But
I may be ignorant of the usage of Dave's package.

Cheers, Ray
-- 
Real programmers confuse Christmas and Halloween because
DEC 25 = OCT 31!

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Raymond Scholz <ray-2001@zonix.de> writes:

> I forgot to mention that I didn't use Mule-UCS before.  With Mule-UCS
> 0.84 utf-8 messages are being created.

Can vanilla Gnus 5.9.0 deal correctly with ISO-8859-15 messages?  Or
do you get multiple parts when you reply to such a message?

Anyway, the current behavior of CVS Gnus + Emacs 21 is not acceptable.
How shall we fix it?  Unify latin-iso8859-1 and latin-iso8859-15
character sets?  Extend the iso-latin-1 and iso-latin-9 coding systems
so that they can encode part of the other charset?  Extend the UTF-8
charset, to make it encode latin-iso8859-15 as well?

The GNU Emacs folks don't seem to be willing to do much work in this
direction because general Unicode support should arrive RSN.

Re: Using iso-8859-15 with Emacs21 & Gnus

[ShengHuo ZHU]

Florian Weimer <fw@deneb.enyo.de> writes:

> Raymond Scholz <ray-2001@zonix.de> writes:
>
>> I forgot to mention that I didn't use Mule-UCS before.  With Mule-UCS
>> 0.84 utf-8 messages are being created.
>
> Can vanilla Gnus 5.9.0 deal correctly with ISO-8859-15 messages?  Or
> do you get multiple parts when you reply to such a message?
>
> Anyway, the current behavior of CVS Gnus + Emacs 21 is not acceptable.
> How shall we fix it?  Unify latin-iso8859-1 and latin-iso8859-15
> character sets?  Extend the iso-latin-1 and iso-latin-9 coding systems
> so that they can encode part of the other charset?  Extend the UTF-8
> charset, to make it encode latin-iso8859-15 as well?

I've put some code into the Gnus CVS to handle this issue.  Please
try. BTW, which charsets are compatible with latin-iso8859-15?

ShengHuo

Re: Using iso-8859-15 with Emacs21 & Gnus

[news]

Florian Weimer <fw@deneb.enyo.de> writes:

> The GNU Emacs folks don't seem to be willing to do much work in this
> direction because general Unicode support should arrive RSN.

Meaning the next release of GNU Emacs?

Chris

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

[-- Attachment #1: Type: text/plain, Size: 550 bytes --]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

>> Anyway, the current behavior of CVS Gnus + Emacs 21 is not acceptable.
>> How shall we fix it?  Unify latin-iso8859-1 and latin-iso8859-15
>> character sets?  Extend the iso-latin-1 and iso-latin-9 coding systems
>> so that they can encode part of the other charset?  Extend the UTF-8
>> charset, to make it encode latin-iso8859-15 as well?
>
> I've put some code into the Gnus CVS to handle this issue.  Please
> try. 

It doesn't work, I'm afraid.  For example, ISO-8859-1 '¼' is changed

[-- Attachment #2: Type: text/plain, Size: 233 bytes --]

to ISO-8859-15 'Œ', which is clearly wrong.

> BTW, which charsets are compatible with latin-iso8859-15?

Strictly speaking, none of the other ISO 8859 ones.  But most
characters (not all) can be encoded using ISO-8859-1, too.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

(Chris Beggy ) news@kippona.com writes:

> Florian Weimer <fw@deneb.enyo.de> writes:
>
>> The GNU Emacs folks don't seem to be willing to do much work in this
>> direction because general Unicode support should arrive RSN.
>
> Meaning the next release of GNU Emacs?

I don't know.  Certainly, this isn't a 21.x issue.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Raymond Scholz]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

> I've put some code into the Gnus CVS to handle this issue.  Please
> try. 

No multipart mails anymore, just text/plain with iso-8859-15.  Great!

> BTW, which charsets are compatible with latin-iso8859-

,----[ man iso-8859-15 ]
| NAME
|        iso_8859-15  -  the  ISO  8859-15 character set encoded in
|        octal, decimal, and hexadecimal
| 
| DESCRIPTION
|        The ISO 8859 standard includes several 8-bit extensions to
|        the  ASCII  character  set  (also  known  as ISO 646-IRV).
|        Especially important is ISO 8859-1,  the  "Latin  Alphabet
|        No.  1",  which  has  become  widely  implemented  and may
|        already  be  seen   as   the   de-facto   standard   ASCII
|        replacement.   However,  it lacks the EURO symbol and does
|        not fully cover Finnish and  French.   ISO  8859-15  is  a
|        modification of ISO 8859-1 that covers these needs.
| 
|        ISO  8859-15  supports  the following languages: Albanian,
|        Basque, Breton, Catalan, Danish, Dutch, English, Estonian,
|        Faroese,   Finnish,  French,  Frisian,  Galician,  German,
|        Greenlandic,  Icelandic,  Irish  Gaelic,  Italian,  Latin,
|        Luxemburgish,   Norwegian,   Portuguese,   Rhaeto-Romanic,
|        Scottish Gaelic, Spanish, and Swedish.
`----

Note that some characters of iso-8859-1 changed their position in
iso-8859-15, so it is not fully backward compatible from this point of
view.

Cheers, Ray
-- 
rscholz:x:587:100:Raymond Scholz:/home/rscholz:/usr/bin/emacs

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Raymond Scholz <ray-2001@zonix.de> writes:

> Note that some characters of iso-8859-1 changed their position in
> iso-8859-15, so it is not fully backward compatible from this point of
> view.

Not quite, they where simply replaced:

-0xA4   0x00A4  #       CURRENCY SIGN
+0xA4   0x20AC  #       EURO SIGN
-0xA6   0x00A6  #       BROKEN BAR
+0xA6   0x0160  #       LATIN CAPITAL LETTER S WITH CARON
-0xA8   0x00A8  #       DIAERESIS
+0xA8   0x0161  #       LATIN SMALL LETTER S WITH CARON
-0xB4   0x00B4  #       ACUTE ACCENT
+0xB4   0x017D  #       LATIN CAPITAL LETTER Z WITH CARON
-0xB8   0x00B8  #       CEDILLA
+0xB8   0x017E  #       LATIN SMALL LETTER Z WITH CARON
-0xBC   0x00BC  #       VULGAR FRACTION ONE QUARTER
-0xBD   0x00BD  #       VULGAR FRACTION ONE HALF
-0xBE   0x00BE  #       VULGAR FRACTION THREE QUARTERS
+0xBC   0x0152  #       LATIN CAPITAL LIGATURE OE
+0xBD   0x0153  #       LATIN SMALL LIGATURE OE
+0xBE   0x0178  #       LATIN CAPITAL LETTER Y WITH DIAERESIS

Re: Using iso-8859-15 with Emacs21 & Gnus

[ShengHuo ZHU]

[-- Attachment #1: Type: text/plain, Size: 595 bytes --]

Florian Weimer <fw@deneb.enyo.de> writes:

> ShengHuo ZHU <zsh@cs.rochester.edu> writes:
>
>>> Anyway, the current behavior of CVS Gnus + Emacs 21 is not acceptable.
>>> How shall we fix it?  Unify latin-iso8859-1 and latin-iso8859-15
>>> character sets?  Extend the iso-latin-1 and iso-latin-9 coding systems
>>> so that they can encode part of the other charset?  Extend the UTF-8
>>> charset, to make it encode latin-iso8859-15 as well?
>>
>> I've put some code into the Gnus CVS to handle this issue.  Please
>> try. 
>
> It doesn't work, I'm afraid.  For example, ISO-8859-1 '¼' is changed

[-- Attachment #2: Type: text/plain, Size: 69 bytes --]

> to ISO-8859-15 'Œ', which is clearly wrong.

Fixed.

ShengHuo

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

[-- Attachment #1: Type: text/plain, Size: 207 bytes --]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

>>> I've put some code into the Gnus CVS to handle this issue.  Please
>>> try. 
>>
>> It doesn't work, I'm afraid.  For example, ISO-8859-1 '¼' is changed

[-- Attachment #2: Type: text/plain, Size: 379 bytes --]

>> to ISO-8859-15 'Œ', which is clearly wrong.
>
> Fixed.

Yes, thanks a lot.  However, I think you should unifiy to ISO-8859-1
instead of ISO-8859-15 because the standard utf-8.el can deal with
ISO-8859-1, but not ISO-8859-15.

BTW, there is a typo in the docstrings for 'mm-hack-charsets', and the
docstring for 'mm-iso-8859-15-compatible' seems to be a bit unclear.

Re: Using iso-8859-15 with Emacs21 & Gnus

[ShengHuo ZHU]

Florian Weimer <fw@deneb.enyo.de> writes:

[...]

> Yes, thanks a lot.  However, I think you should unifiy to ISO-8859-1
> instead of ISO-8859-15

I think that in most cases ISO-8859-15 is used because of the euro
sign, which is not convertible to ISO-8859-1.

> because the standard utf-8.el can deal with ISO-8859-1, but not
> ISO-8859-15.

That's the problem of utf-8.el.

> BTW, there is a typo in the docstrings for 'mm-hack-charsets', and the
> docstring for 'mm-iso-8859-15-compatible' seems to be a bit unclear.

Fixed.

ShengHuo

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

>> because the standard utf-8.el can deal with ISO-8859-1, but not
>> ISO-8859-15.
>
> That's the problem of utf-8.el.

It completely breaks UTF-8 support for Gnus using Emacs 21.  If you
reply to a UTF-8 message, chances are good that it contains ISO-8859-1
characters, and if I understand your code correctly, these characters
are mapped to ISO-8859-15, so the reply cannot be encoded using UTF-8
by Emacs.

But I agree that utf-8.el should support more charsets.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Per Abrahamsen]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

> I think that in most cases ISO-8859-15 is used because of the euro
> sign, which is not convertible to ISO-8859-1.

I might misunderstand you, but ISO-8859-1.is used a lot more than
ISO-8859-15 right now.  It is not clear that this will change in the
future, since use is likely to migrate to UTF8 rather than
ISO-8859-15. 

Re: Using iso-8859-15 with Emacs21 & Gnus

[ShengHuo ZHU]

Florian Weimer <fw@deneb.enyo.de> writes:

> ShengHuo ZHU <zsh@cs.rochester.edu> writes:
>
>>> because the standard utf-8.el can deal with ISO-8859-1, but not
>>> ISO-8859-15.
>>
>> That's the problem of utf-8.el.
>
> It completely breaks UTF-8 support for Gnus using Emacs 21.  If you
> reply to a UTF-8 message, chances are good that it contains ISO-8859-1
> characters, and if I understand your code correctly, these characters
> are mapped to ISO-8859-15, so the reply cannot be encoded using UTF-8
> by Emacs.

The point is that if a message contains ISO-8859-15, Gnus converts
ISO-8859-1 to ISO-8859-15.

The cases are that if a message contains

1. no ISO-8859-15 chars, no conversion. Fine for utf-8.el.

2. some inconvertible ISO-8859-15 chars (e.g. euro sign), convert
   ISO-8859-1 to ISO-8859-15. No harm, utf-8.el can't encode it either
   way.  

3. no inconvertible ISO-8859-15 chars, convert ISO-8859-1 to
   ISO-8859-15. It would be better to convert them to ISO-8859-1. But
   this case doesn't happen much. Why users input ISO-8859-15 chars if
   they can do it with ISO-8859-1 chars?

To autodetect the case 3 will lose the efficiency.

ShengHuo

Re: Using iso-8859-15 with Emacs21 & Gnus

[Kai Großjohann]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

> 3. no inconvertible ISO-8859-15 chars, convert ISO-8859-1 to
>    ISO-8859-15. It would be better to convert them to ISO-8859-1. But
>    this case doesn't happen much. Why users input ISO-8859-15 chars if
>    they can do it with ISO-8859-1 chars?

Presumably, users will switch from a _default_ encoding of Latin-1
now to a default encoding of Latin-9 later.  Most people probably
don't want to use Latin-1 for most things, and Latin-9 only for the
Euro sign -- they want to use Latin-9 in all cases.

At least that's what I would want, as a (more or less) typical German.

kai
-- 
Lisp is kinda like tpircstsoP

Re: Using iso-8859-15 with Emacs21 & Gnus

[Graham Murray]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> Presumably, users will switch from a _default_ encoding of Latin-1
> now to a default encoding of Latin-9 later.  Most people probably
> don't want to use Latin-1 for most things, and Latin-9 only for the
> Euro sign -- they want to use Latin-9 in all cases.

Though does USEFOR not specify a default of utf-8?

Re: Using iso-8859-15 with Emacs21 & Gnus

[Karl Eichwalder]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

> 3. no inconvertible ISO-8859-15 chars, convert ISO-8859-1 to
>    ISO-8859-15. It would be better to convert them to ISO-8859-1. But
>    this case doesn't happen much. Why users input ISO-8859-15 chars if
>    they can do it with ISO-8859-1 chars?

Happens automatically when you call Emacs from an ISO-8859-15 locale:

    LANG=de_DE.ISO-8859-15  emacs

or

    LANG=de_DE@euro  emacs

-- 
ke@suse.de (work) / keichwa@gmx.net (home):              |
http://www.suse.de/~ke/                                  |      ,__o
Free Translation Project:                                |    _-\_<,
http://www.iro.umontreal.ca/contrib/po/HTML/             |   (*)/'(*)

Re: Using iso-8859-15 with Emacs21 & Gnus

[Per Abrahamsen]

Florian Weimer <fw@deneb.enyo.de> writes:

> Graham Murray <graham@barnowl.demon.co.uk> writes:
>
>> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>>
>>> Presumably, users will switch from a _default_ encoding of Latin-1
>>> now to a default encoding of Latin-9 later.  Most people probably
>>> don't want to use Latin-1 for most things, and Latin-9 only for the
>>> Euro sign -- they want to use Latin-9 in all cases.
>>
>> Though does USEFOR not specify a default of utf-8?
>
> No.  USEFOR does not specify any default charset.  

Yes it does.  UTF-8 for headers and US-ASCII for the body.

> Anyway, I think Kai was talking about the default charset for posting,
> which is properly MIME-declared by Gnus.

Agreed.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Graham Murray <graham@barnowl.demon.co.uk> writes:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>
>> Presumably, users will switch from a _default_ encoding of Latin-1
>> now to a default encoding of Latin-9 later.  Most people probably
>> don't want to use Latin-1 for most things, and Latin-9 only for the
>> Euro sign -- they want to use Latin-9 in all cases.
>
> Though does USEFOR not specify a default of utf-8?

No.  USEFOR does not specify any default charset.  It doesn't even
encourage reading agents to implement hierarchy-specific default
charsets, which is required in practice.

Anyway, I think Kai was talking about the default charset for posting,
which is properly MIME-declared by Gnus.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Russ Allbery]

Per Abrahamsen <abraham@dina.kvl.dk> writes:
> Florian Weimer <fw@deneb.enyo.de> writes:

>> No.  USEFOR does not specify any default charset.  

> Yes it does.  UTF-8 for headers and US-ASCII for the body.

Why on earth would you require UTF-8 handling for the headers and not take
advantage of it for the body?  Oh, wait, right, because we're rewriting
the mail standards without really thinking.

It's this kind of thing that really makes me think it's best to ignore
USEFOR completely for right now.  The lunatics are running the asylum over
there.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

USEFOR (was: Re: Using iso-8859-15 with Emacs21 & Gnus)

[Per Abrahamsen]

Russ Allbery <rra@stanford.edu> writes:

> Per Abrahamsen <abraham@dina.kvl.dk> writes:
>> Florian Weimer <fw@deneb.enyo.de> writes:
>
>>> No.  USEFOR does not specify any default charset.  
>
>> Yes it does.  UTF-8 for headers and US-ASCII for the body.
>
> Why on earth would you require UTF-8 handling for the headers and not take
> advantage of it for the body?

Because leaving out the Content-Type header for messages with UTF-8
content in the body cause problems for pre-standard clients and has no
real advantages.

The situation is very different for headers, where both the option of
qp-encoding UTF-8 content in the headers and the option of not
qp-encoding UTF-8 content in the header will cause real problems for
pre-standard newsreaders.  And the option of qp-encoding the header
will cause problems on pre-standard servers.

The cases are not parallel, since MIME does not allow you to specify
character set in headers without QP or BASE64 encoding the content.
In the body, the two functions are ortogonal.

I agree that USEFOR in general are doing too much design and too
little standardization, but I think they took the least bad option in
this case.

Re: USEFOR (was: Re: Using iso-8859-15 with Emacs21 & Gnus)

[Russ Allbery]

Per Abrahamsen <abraham@dina.kvl.dk> writes:
> Russ Allbery <rra@stanford.edu> writes:

>> Why on earth would you require UTF-8 handling for the headers and not
>> take advantage of it for the body?

> Because leaving out the Content-Type header for messages with UTF-8
> content in the body cause problems for pre-standard clients and has no
> real advantages.

Ah, yeah, good point.

> I agree that USEFOR in general are doing too much design and too little
> standardization, but I think they took the least bad option in this
> case.

Yeah, you're right, they did.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

>> No.  USEFOR does not specify any default charset.  
>
> Yes it does.  UTF-8 for headers and US-ASCII for the body.

For the headers, I agree, but for the body, no.  There's no such
default in the current draft.

Re: USEFOR

[Florian Weimer]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> Because leaving out the Content-Type header for messages with UTF-8
> content in the body cause problems for pre-standard clients and has no
> real advantages.

In particular since many hierarchies have already a declared a charset
to be the default if no MIME headers are present.  Unfortunately,
USEFOR is completely ignoring this issue and fails to mention that
reading agents need hierarchy-specific charset defaults.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Russ Allbery]

Florian Weimer <fw@deneb.enyo.de> writes:

> For the headers, I agree, but for the body, no.  There's no such default
> in the current draft.

It's inherited from MIME.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Using iso-8859-15 with Emacs21 & Gnus

[Paul Jarc]

Per Abrahamsen <abraham@dina.kvl.dk> wrote:
> ShengHuo ZHU <zsh@cs.rochester.edu> writes:
>> I think that in most cases ISO-8859-15 is used because of the euro
>> sign, which is not convertible to ISO-8859-1.
>
> I might misunderstand you, but ISO-8859-1.is used a lot more than
> ISO-8859-15 right now.

I think ShengHuo meant: in most of the cases where -15 is used, it is
used for the euro sign; not: -15 is used more often than -1.

So unifying -15 and -1 to -1 is likely to cause lossage: the original
-15 part probably used to euro, and the unified message won't be able
to.


paul

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Russ Allbery <rra@stanford.edu> writes:

> Florian Weimer <fw@deneb.enyo.de> writes:
>
>> For the headers, I agree, but for the body, no.  There's no such default
>> in the current draft.
>
> It's inherited from MIME.

How do you know? ;-)

The draft encourages implementations to recognize UTF-8 even though it
is not declared (and fails to mention the already existing practice of
hierarchy-specific default charsets).

Re: Using iso-8859-15 with Emacs21 & Gnus

[Russ Allbery]

Florian Weimer <fw@deneb.enyo.de> writes:
> Russ Allbery <rra@stanford.edu> writes:
>> Florian Weimer <fw@deneb.enyo.de> writes:

>>> For the headers, I agree, but for the body, no.  There's no such
>>> default in the current draft.

>> It's inherited from MIME.

> How do you know? ;-)

How do I know which?  That news messages are supposed to be MIME messages?
The draft says that pretty clearly, I thought, although it then mentions a
bunch of exceptions like UTF-8 headers.

> The draft encourages implementations to recognize UTF-8 even though it
> is not declared (and fails to mention the already existing practice of
> hierarchy-specific default charsets).

The draft completely ignores all sorts of existing practice that the
authors seem to find inconvenient.  *sigh*

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Using iso-8859-15 with Emacs21 & Gnus

[Kai Großjohann]

Graham Murray <graham@barnowl.demon.co.uk> writes:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>
>> Presumably, users will switch from a _default_ encoding of Latin-1
>> now to a default encoding of Latin-9 later.  Most people probably
>> don't want to use Latin-1 for most things, and Latin-9 only for the
>> Euro sign -- they want to use Latin-9 in all cases.
>
> Though does USEFOR not specify a default of utf-8?

I meant that I now use (set-language-environment "Latin-1"), and once
I see a real need for the Euro sign I might switch to
(set-language-environment "Latin-9"), instead.

This change affects all of Emacs, so USEFOR is only marginally
relevant, here.

kai
-- 
I like BOTH kinds of music.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Kai Großjohann]

Florian Weimer <fw@deneb.enyo.de> writes:

> Anyway, I think Kai was talking about the default charset for posting,
> which is properly MIME-declared by Gnus.

Well, err.  Hm.  I wasn't thinking of this explicitly, but in effect,
that's what I was saying: right now, Gnus uses Latin-1 for German
Umlauts, and after I switch to Latin-9, Gnus should use Latin-9 for
the Umlauts, too.

In the current design of Mule, this would be achieved by the
following chain of events:

* I change (set-language-environment "Latin-1") to
  (set-language-environment "Latin-9").

* Now typing `" a' inserts a Latin-9 ä which Emacs distinguishes from
  a Latin-1 ä.

* Gnus sees non-Ascii Latin-9 characters in my post and chooses
  iso-8859-15 as charset.

After changing Mule to unify the various Latin-* charsets, Emacs does
not distinguish the two variants of ä anymore.  Not sure what that
means to the whole procedure.  Maybe Latin-9 is preferred and used
whenever possible, after set-language-environment?  Maybe.

kai
-- 
I like BOTH kinds of music.

Re: Using iso-8859-15 with Emacs21 & Gnus

[Florian Weimer]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> After changing Mule to unify the various Latin-* charsets, Emacs does
> not distinguish the two variants of ä anymore.  Not sure what that
> means to the whole procedure.  Maybe Latin-9 is preferred and used
> whenever possible, after set-language-environment?  Maybe.

If unification is present, the whole MIME charset decision process in
Gnus has to be redesigned (and many things can then work in a very
straightforward way).

Re: USEFOR

[Per Abrahamsen]

Florian Weimer <fw@deneb.enyo.de> writes:

> In particular since many hierarchies have already a declared a charset
> to be the default if no MIME headers are present.  Unfortunately,
> USEFOR is completely ignoring this issue and fails to mention that
> reading agents need hierarchy-specific charset defaults.

They don't.  It is a quality of implementation issue.