The simplest model is that we suppose we should return the japanese message when the ML is configured as for Japanese. This model needs that the language of a ML is fixed. This ML always returns the error message in Japanese.
However, it is not proper to ignore Accept-Language: header field since some Japanese people cannot read Japanese on the screen. There may be invalid Accept-Language: field. There are several problems.
For example, if "Accept-Language: ja" is given, we can recognize the language is "ja". However, how about the case Japanese send a mail with the body content "help" (English) ? In that case we cannot determine we should return the reply in us-ascii or iso-2022-jp. For example, how about the following example ?
[Example] From: email@example.com Subject: help mime-version: 1.0 content-type: text/plain; charset=us-ascii help
We should return the English reply message based on charset information since this message contains English only, so charset is us-ascii. But in this case we should return Japanese message since the sender is a Japanese. So we should return the help message in both English and Japanese if the ML is configured as Japanese preferrable. If English is preferrable, the reply is English only.
If Accept-Lanaguage: is specified, we prefer it. Even if the ML is English preferrable (config: language_preference_order = en) and the message with "Accept-Lanaguage: ja" is given, the ML returns the Japanese help message. fml8 behaves like it.
Copyright (C) 1993-2022 Ken'ichi Fukamachi mail:< fukachan at fml.org >