It's the 21st Century. We should be using UTF-8. But when I compose a mail in mutt using non-ascii characters, it'll send it as iso8859-1 (unless I force the issue by using characters not in iso8859-1 either).
How is set the send_charset variable?
That would be in .muttrc? I don't have a .muttrc -- I only fired up mutt to reproduce this bug when it was reported to me.
Setting it to 'us-ascii:utf-8' in /etc/Muttrc fixes the problem. Please consider doing this in our package (and likewise for any other cases where we might use legacy 8-bit charsets instead of utf-8). Thanks.
What exactly is the problem caused by using iso-8859-1? "us-ascii:iso-8859-1:utf-8" is the upstream default and I'm not sure there isn't a client that can handle iso-8859-1, but not utf-8.
iso8859-1 is a legacy character set. One of the problems with charset labelling is that the labels tend to "fall off". If we were good at labelling everything and keeping the labels, things would have been a lot easier and a lot of the motivation for UTF-8 wouldn't have been there. One of the main advantages of our "UTF-8 everywhere" policy is that the labelling increasingly doesn't matter -- it's more and more valid just to assume UTF-8. Anything generated within your own system (be that a single computer, a network of computers, whatever) should be in UTF-8. In this particular case, for example, I should be able to grep my local maildir of outgoing mail for 'naïve' and get the expected result. I wouldn't expect to find my own outgoing mail encoded in iso8859-1, euc-jp, or anything but UTF-8 (or ascii, as a subset of UTF-8). Another case where it causes problems is sending patches for the Linux kernel. The kernel is in UTF-8, and patches which are charset-converted will either not apply, or will add invalid text to the kernel (git-am will do charset conversion from legacy charsets on the cosmetic parts, including the commit comment, but it leaves the body of the patch exactly as it was).
Ok, I'll change it in the next release. Thanks.
I think that using iso-latin1 when there is only iso-latin1 characters is safer. On the receiving end the system may be much older than fedora and don't know about utf8, and though some systems may not know about utf8 all should know about latin1.
It seems that thunderbird is also using iso8859-1 by default, so mutt isn't the only MUA in Fedora behaving like this. I'll leave it as it is for now. Users that need utf8 to be used can always override the setting in .muttrc.