[rt-devel] Re: [PATCH] Preserve original charset + fix bogus UTF32LE guess on binary attachments

Dmitry Sivachenko mitya at cavia.pp.ru
Sat Jun 28 05:25:05 EDT 2003


On Sat, Jun 28, 2003 at 05:22:11PM +0800, Autrijus Tang wrote:
> On Sat, Jun 28, 2003 at 11:09:47AM +0400, Dmitry Sivachenko wrote:
> > On Sat, Jun 28, 2003 at 01:39:00AM +0800, Autrijus Tang wrote:
> > > This is not something to worry about at all.  "Detecting" including
> > > using the explicit info.  It's not "guessing".
> > 
> > Well, I was talking about _GuessCharset subroutine, which only *guesses*
> > charset via Encode::Guess.  I see no reason to guess something in
> > http://mitya.pp.ru/rt-bad , because all non-ascii characters it in are
> > properly encoded.
> > Probably I am missing something.
> 
> Yeah.  RT still needs to Guess the charset for the main entity,
> because it says:
> 
>     Content-Type: multipart/mixed;
> 	    boundary="----=_NextPart_000_0FF2_01C33CAF.BB8C8E00"
> 
> and mentions nothing about its own Content-Type.Charset.  Hence in
> order to determine how to decode its headers, like Subject:
> 
>     Subject: =?windows-1251?B?ze7i7vHy6CDu8iDI7fLl6y3S?=
> 
> RT has to feed the headers to _GuessCharset.  Since it contains
> only ascii characters, 'ascii' will be used and no transformation
> will take place.
> 
> This is merely a safeguard against broken MUAs that sends out
> headers that contain high-bit characters.

Thank you for explanation!

> 
> > > Can you try the following patch against RT::Action::SendEmail
> > > and see if it works?
> > 
> > Yes, not filename of the attachment is properly encoded, thanks!
> 
> I assume s/not/now/.  Good then! :-)

Yes, sure!



More information about the Rt-devel mailing list