[Mailman-Developers] Mailman CVS sends outJapanesetemplatemails in EUC-JP

Wed, 12 Sep 2001 14:32:12 +0900

>>>>> "Tokio" == Tokio Kikuchi <tkikuchi@is.kochi-u.ac.jp> writes:

    Ben> I did this, and welcome mails are still mojibake..

    Tokio> Have you restart the qrunner by bin/mailmanctl stop
    Tokio> bin/mailmanctl start I have not tried whether restart works
    Tokio> or not, yet.

Wow!! I feel very silly, I didn't even think to try that.

Now, with the settings:

DEFAULT_SERVER_LANGUAGE = 'ja'
GLOBAL_PIPELINE = [ 'ja_To_EUC_JP', ] + GLOBAL_PIPELINE
DELIVERY_MODULE = 'ja_SMTPDirect'

in Mailman/mm_cfg.py, welcome mails are sent out properly encoded,
subjects and all.

Pipermail.py is still not filling out the %(encoding) parameter
kof index_header_template:

index_header_template='''<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
  <HEAD>
     <title>The %(listname)s %(archive)s Archive by %(archtype)s</title>
     <META NAME="robots" CONTENT="noindex,follow">
     %(encoding)s
  </HEAD>

[snip]

You can see how it's coming out on my test archive page:

http://nausicaa.interq.or.jp/pipermail/test2/2001-September/000000.html

The mail is mostly converted to EUC-JP, and the rest of the page is
EUC, but the %(encoding) parameter is left blank, so my web browser
assumes it's iso-8859-1.

The hankaku (half-width) katakana characters in the message in the
above URL are not converted to EUC-JP correctly, and are mojibake.

Although the kconv documentation at
http://tomigaya.shibuya.tokyo.jp/~mak/kconv/readmeen.txt says
that the default is to convert hankaku katakana to full-width
zenkaku katakana:

    * hankanaconvert

        Specify whether kconv converts hankaku katakana to zenkaku
        katakana or not. If the output encoding is EUC and you do want
        not to convert hankaku katakna to zenkaku katakana, kconv
        outputs 2 byte hankaku katakana in EUC.

        kconv.ZENKAKU - Convert hankaku katakana to zenkaku
        katakana(DEFAULT)

        kconv.HANKAKU - Output hankaku katakana if input string
        includes hankaku katakana

the above URL shows that the hankaku katakana are *NOT* being
converted to zenkaku.  

In fact, I just tried changing ja_To_EUC_JP.py and
ja_To_ISO_2022_JP.py to specify kconv.HANKAKU, sent a test
message with hankaku katakana, and still got mojibake in the
archives.

Finally, I tried specifying kconv.ZENKAKU, to see if the hankaku would
be converted, and they were not.  You can see the results of all these
tests at:

http://nausicaa.interq.or.jp/pipermail/test2/2001-September/

I think kconv has some bugs with converting ISO-2022-JP to EUC when
hankaku characters are involved.  Have you seen this problem,
Kikuchi-san?

Ben

-- 
Brought to you by the letters S and P and the number 19.
"Frungy! Frungy! Frungy!!"
Debian GNU/Linux maintainer of Gimp and GTK+ -- http://www.debian.org/