> I feel the messages file should contain exactly what came down the wire.

The messages file currently does *not* have what came down the wire --
Internet standard text uses CRLF terminaters, while Squeak uses CR-only.
 We've already made one concession towards convenience, and we might as
well continue.

The one interesting facet is that there is some hope that Squeak will
switch to something like Latin-1 in the future.  If so, then we would
have to convert message files *back* again.  However, if my hope is
misplaced, and Squeak is really going to stick with the Mac encoding for
a year or longer, then I still think the message files should be 

The reason has nothing to do with the encoding's complexity or with the
time.  After all, #squeakToIso and its complement are already written. 
Furthermore, the cost of reading the message from disk likely dwarfs the
cost of the translation.  The reason I'd like the messages file in the
same thing as Squeak's regular String format is that the code doesn't
have to keep track of what kind of String's it is using -- $a is $a, and
umlaut is umlaut.

Regarding index files, I like the decoding of MIME headers in advance,
for the same reason: solve the problem once and then stop worrying about
it.  If that is what Mike's patch does, then that is excellent.

By the way, is anyone testing this stuff on Filter Celeste?


-Lex