On Wed, Dec 14, 2011, at 11:21, Stefan Sperling wrote: > On Wed, Dec 14, 2011 at 12:10:25PM +0200, Daniel Shahaf wrote: > > Andreas' attachment declared itself as > > Content-Type: text/x-patch; charset=us-ascii; name="german-fuzzy-0.diff" > > but contained literal [ü»«] characters. > > > > Andreas, you should consider fixing that, so it's easier to process > > your patches. > > This could also be an interoperability problem between mutt and gmx. > The charset is correct as far as the encoded base64 form is concerned. > Should the charset= header apply to the encoded or the decoded form? > Of course, always assuming ascii for the encoded form makes sense for > base64. But what about other encodings? Not sure what RFCs say here. >
AIUI: The charset= applies to the decoded form. If you were to introduce base128 which encodes arbitrary files into an 8-bit encoding, you'd use Content-Transfer-Encoding: 8bit on the message as a whole (and Content-Transfer-Encoding: base128 on the attachment MIME part). In particular: consider the case of a binary attachment. It won't be text/plain and hence won't have a charset= parameter. But the MIME headers still have to say how it's encoded (literally, base64, quoted- printable)
