Re: Trouble displaying 8bit postings properly

[Jesper Harder <harder@myrealbox.com>]

Karl Eichwalder <keichwa@gmx.net> writes:

> Can you still access this posting?
>
>     Message-ID: <3E7260DE.77CC579B@gmx.de>
>
> Emacs started in an UTF-8 locale (LANG=de_DE.UTF-8 emacs), failed to
> display the umplauts properly; the umlauts are displayed correctly when
> I type 'C-u g'
>
> The same is to be observed for this article (where '1 g ISO-8859-1 RET'
> fails):

You can do `C-u W M c latin-1' instead.

>     Message-ID: <3e73794b.14305425@news.cis.dfn.de>

The problem in both cases is that the articles contain octets that are
invalid in Latin-1.  

1. The first seems to be slightly corrupted for some reason -- notice the
   line ending with "Preissystem begr\201át".

2. The second is actually encoded in the evil Windows-1252 charset --
   the hyphen in "Hamburg-Rostock" is an n-dash.

The real solution to 2. is to have Emacs know about Windows-1252, which
it doesn't.  I think it's supported in the Unicode branch, though.

The first case is more difficult.  It's a tradeoff -- it would have
displayed correctly with the old code which doesn't try guess the right
encoding.  On the other hand, all latin-1 articles without a MIME
charset would be displayed incorrectly with the old code.

I think valid latin-1 articles without a charset declaration are more
common than than articles with a strange corruption -- so it's probably
a reasonable tradeoff [1].  What do you think?

[1] But note that these problems only apply in an UTF-8 locale -- both
articles display correctly in a latin-1 locale (except for the invalid
octets, of course).