Santiago Garcia Mantinan wrote:

Suppose I have a list which default language is using charset iso-8859-1, then somebody comes and posts in iso-8859-15 and you get it all in three mime parts, one for the header, in iso-8859-1, one for the body in iso-8859-15 and then the footer in iso-8859-1 again.

This is a big problem, one I didn't address with my initial charset support implementation for Mailman. I'll add this functionality as soon as I can, hopefully tonight or tomorrow. Sorry for all the inconvenience. Basically, we need to have Mailman's Charset.py treat iso-8859-1 exactly the same as iso-8859-15 as a special case. Ben

On Wed, 2003-10-29 at 17:27, Martin v. Löwis wrote:

If you want to add it, please look at Mailman/Handlers/Decorate.py. The test

if not msg.is_multipart() and msgtype == 'text/plain' and \ msg.get('content-transfer-encoding', '').lower() <> 'base64' and \ (lcset == 'us-ascii' or mcset == lcset):

is the one that allows direct concatenation of the header. You could extend this to recode the header into the message charset.

Are Latin 1 and Latin 9 essentially compatible? Would something like the following work: def compatible(mcset, lcset): # If the list's preferred charset is us-ascii, we can always safely add # the header/footer to a plain text message since all email charsets # Mailman supports are strict supersets of us-ascii -- no, UTF-16 emails # are not supported. if lcset == 'us-ascii': return True if mcset == lcset: return True # Latin 1 and Latin 9 are basically compatible if lcset in LATIN_1_15 and mcset in LATIN_1_15: return True # For now, nothing else is compatible. return False ...and then replace the last conditional in the above with: compatible(mcset, lcset) ? -Barry

Barry Warsaw writes:

Are Latin 1 and Latin 9 essentially compatible? Would something like the following work:

It would be too dangerous. It *would* help if you could try to perform the recoding of the actual data. Regards, Martin

Hi, Folks. Barry Warsaw wrote:

On Mon, 2003-12-01 at 02:15, Martin v. Löwis wrote:

...

Barry Warsaw writes:

...

Are Latin 1 and Latin 9 essentially compatible? Would something like the following work:

It would be too dangerous. It *would* help if you could try to perform the recoding of the actual data.

Which would be too much for a patch release I think. Okay, thanks all, I'll punt on this one for 2.1.4.

Do you remember this? I have uploaded a patch to improve the i18n feature of mailman at http://sourceforge.net/tracker/index.php?func=detail&aid=865661&group_id=103&atid=300103 This includes a patch for Decorate.py to do the best to keep plain text message as not being multiparted. The strategy is; 1. First convert the header, body and footer into unicode. 2. Then try to encode it with the list charset. 3. If it fails, the message body may have contained characters like euro sign. So, try to encode it with the message charset. 4. If both fails, fall back to the multipart message. If you keep the footer only in us-ascii for French or German list, you will always get non-multipart message even if the users post with euro sign. Cheers, -- Tokio Kikuchi, tkikuchi@ is.kochi-u.ac.jp http://weather.is.kochi-u.ac.jp/