From: (no name) (no email)
Date: Thu Feb 26 2004 - 15:27:21 EST
On Thu, 26 Feb 2004, Noel Jones wrote:
> This was discussed on the anomy-users list a couple months ago.
> For unknown reasons (some version of?) Outlook insists on sending PDF files
> as QP instead of Base64. Since anomy unpacks and repacks all attachments,
> any "Newline" got translated to LF, where windows was expecting it to be CRLF.
> Such PDFs were then unusable by windows users after passing through anomy.
> I think the ugly workaround anomy came up with was to always use CRLF when
> encoding QP segments, knowing that this might break unix and mac QP
> attachments.
>
Actually this is false. SMTP and MIME define a canonical representation
for text/* entities which is CRLF separated. When using a non identity
transfer encoding (not "7bit", "8bit" or "binary") all MIME Entities need
to be first converted to their canonical representation. When decoding
text/* entities any resulting CRLF should be converted to the local text
representation. Any (encoded) line breaks in a QP encoded text/* Entity
MUST be encoded as "=0D=0A" and this does break compatibility with UNIX or
MacOS, any MIME decoder that does QP decoding without paying attention to
the Content-Type is broken.
-- Viktor.
|
|
|