Kai Grossjohann <grossjoh@ls6.informatik.uni-dortmund.de> writes:

> I think this problem is AI complete if you want to do it properly. 

Well, I was planning on settling for non-proper solutions, then.

> The author of a mail message wrote it in a monospaced (as opposed to
> proportional) font.  The author of a message tries to make it
> readable.

...but then he shouldn't use a monospaced font! :-)

Besides, that's not really the issue -- the issue is when to break
lines, and when not to.  It's OK to accept things as they come on a 80
column vt100, but I have a fairly large Emacs window, and I'd like to
have text something more than just a narrow column.

I understand that it might be difficult to recognize when things can't
be broken, but some heuristics I'd suggest would be:

o lines with very few (non-wsp) characters, say less than 30
o lines with lots of non-letters [|\(){},+-*/&%$#]
o lines between «begin 644» and «end» :-)

Perhaps what we really need is for some well chosen SGML tags to get
accepted on Usenet.  How about an SGML-hierarchy?  RFD sgml.general,
anyone?  Have tags like <code>, <text>, <sig>, <footnote>, for instance.

> I'd imagine recognizing C might be rather easy: if the lines ends in
> ";" it's C, if the prev and/or the next line is C and the line
> contains "else", it's C, too.

And wouldn't it be way cool to have this display as any other C-code
(read: cc-mode/font-lock'ed)?  I likes it.  I likes it a lot.

> But how would you tell the difference between English and Cobol in
> "add a to b giving c"?

..and I'll worry about _that_ when somebody sends me messages with cobol
in them.  (Hey, how's font-locking in cobol-mode, anyway? :-)

~kzm
--
    Mail from aol.com and interramp.com domains will be discarded
          Finger <ketil@haukugle.ii.uib.no> for more info