Re: [tlug] Redhat Enterprise and PHPs MBSTRING

[David E <dave@???>]

From: Robert Fisher <robert@example.com>
Subject: Re: [tlug] Redhat Enterprise and PHPs MBSTRING
Date: Thu, 31 Mar 2005 19:10:17 +0900

> David E wrote:
> >From: Robert Fisher <robert@example.com>
> >>David E wrote:
> >>>From: Robert Fisher <robert@example.com>
> >>>
> >>>>I am installing a system and for some reason we are having a problem 
> >>>>with encoding of japanese text to be sent out in emails.  The emails 
> >>>>either come in "mojibake" or cutoff.  
> >>>Have you checked the mb-related settings in the two php.ini files, and/or
> >>>compared the output of phpinfo() on the 2 systems?
> >>>
> >>Thanks for your response, yeah was the very first thing we did.  
> >>
> >>Would the Locale of the server have any effect?  on both servers the 
> >>locale is english utf8.
> >>    
> >I'd still be inclined to look for some difference between the two
> >systems in one of: the php code, php installation settings, php.ini settings.
> >
> >As in Alessandro's example, is mb_language set to "ja" in both? Are
> >both using mb_send_mail rather than send_mail? That sort of thing. 
> >
> 
> mb_language is set to "ja" on both servers. and both servers are using 
> mb_send_mail.
> 
> I am actually trying to figure out why it is so different between the 
> two servers.  We did however come up with a solution.  On the red hat 9 
> server, we converted the body of the email to ISO-2022-JP  and the 
> subject line was in EUC format before passing it through mb_send_mail.  
> On the enterprise server, the solutions was to send the subject and body 
> into mb_send_mail as EUC formatted.  I understand that mb_send_mail will 
> convert the text to ISO-2022-JP format. 
> 

I'd guess that passing in the subject and body in different encodings
could cause problems with the conversion. When mb tries to detect the
input encoding, it may just go with the first one it detects (either
the subject or the body) and then convert everything based on
that. 

Since you know what encoding you're starting with, why not expressly
convert both the body and subject to ISO-2022-JP before passing them
to mb_send_mail and avoid the risk that mb's auto-detect might get it
wrong?

-dave