Lex,

My approach in this change set was to try to keep things as simple as
possible, while not touching the messages file data at all.

What is stored in the messages file is exactly what came across the
wire.  That means that each time it is displayed, it is converted using
isoToSqueak (though this is not very expensive).

What is stored in the index file is (if needed) MIME decoded and then
converted using isoToSqueak.  That means you can filter for a subject
field containing "joe", whether it was originally in plain text or was
MIME encoded.  Previously MIME headers were effectively unfilterable.

The good news is that a lot of the old randomly scattered conversions
are now a lot more centralized.  All of the work to generate index file
entries is done in one method (IndexFileEntry readFrom: aStream)

I hope this description helps.  I think the change set makes the
situation a lot more clear and simple than it was before.

Mike