Re: [Nmh-workers] charsets

[Ken Hornstein]

>1) on input (inc, slocal, etc), we convert to UTF-8 [1]

So, I suggested that a while ago .... and I got pushback.

http://lists.nongnu.org/archive/html/nmh-workers/2012-01/msg00206.html

>2) on output, we convert to whatever the environment wants.
>
>3) Everything inside the code/disk-storage uses UTF-8.

I'm not such in favor of that part.  I'm probably going to implement
better multipart/signed and multipart/encrypted support, and keeping
around the original message would be a requirement for that.  Also, it
just offends my sensibilities to mangle messages after we receive them,
but I recognize that isn't exactly a valid technical argument.  But ...
it's not clear to me what inc & slocal should do if they get a malformed
MIME message; right now since those utilties don't have to parse any
MIME content they can write out a malformed message and the user has a
chance to deal with it.

>Ditch stdio in favour of bio and we are 90% of the way
>there. Incorporate bio in the build tree and we have no external
>dependencies.
>
>[1] http://plan9.bell-labs.com/magic/man2html/2/bio

Hm.  It's not clear to me how much bio helps here.  We don't ...
in general ... care about runes.  The characters it cares about for
parsing & the like are ASCII.  Everything else is basically an opaque
blob.  Converting from q-b, base64, and non-native character sets are
what we really care about.

--Ken