TODO

Integrate --debug-imap option into yet-to-be-implemented -vv switch? 
I had the idea to provide separate debugging info levels anyway, see --debug
below.

Gracefully close IMAP connection upon unexptected error (currently archivemail
just terminates). 

LOCKING & Co:
* Block signals while writing changed mailbox back.
* Double-check the entire locking code. 

Seems like existing archives are not read or validated in any way.  New archive
data is blindly appended...  Probably okay, but should be documented. 

IMAP SEARCH BEFORE disregards time and timezone information.  This should at
least be documented.  E.g. I've found that '-d 0' didn't match all messages in
an IMAP mailbox.  This is because the SEARCH key is (BEFORE 14-Nov-2007) on 14
November, not matching messages that arrived today.  (This problem is probably 
fixed for most use cases by the --all option.) 

Document mbox format issues: link to
http://homepages.tesco.net/~J.deBoynePollard/FGA/mail-mbox-formats.html, 
qmail mbox manpage, Debian manpage, RFC 4155.  Document what mbox format we can
read, and what we write. 
FIXME: we cannot yet parse rfc 2822 addr-spec stuff like quoted local-parts in
return-path addresses.

Minor annoyance: when a From_ line is generated, guess_delivery_time() reports
the used date header a second time. 

Check sf.net and Debian BTS for new bugs.  Again.

IMAP: ensure mailbox archives are properly named.  Currently imap folder names
are mapped like this:

  IMAP URL    |  resulting mbox_archive
  ------------+------------------------
  test.box    |  test.box_archive.gz
  test/box    |  box_archive.gz


Implement --include-draft.  But before, think about it again.  (This is feature
request #1569305.)

Implement a fallback if an IMAP server doesn't support SEARCH. (Ouch!)

Add IMAP tests to the testsuite (upload test messages with IMAP "APPEND
date-string").  This should be done without any real network I/O.

Try to port archivemail to email.message and the new mailboxes in Python 2.5.
Is these flexible enough for our needs?

Line out what we want with respect to multiple selection criteria. 
Some make sense, but this easily gets too complex, and if only it's a hassle
with adding all the options.  Hm.  

Reject patch #1036022 "Added option to inverse date compare" after cooling down
because the patch is both stupid (copy+paste code) and broken.  Don't see why
anyone should want this/we should support it. 
If this is reasonable *at all*, I think we'd better go for all the complexity
to honour _two_ cut off dates (see Debian bug "#184124: archivemail: -D and -d
should not be incompatible", which is a comparably half-baken thought). </rant>

Add --debug or -vv switch, and move the printing of diagnostic info for each
message to --debug.  

Perhaps add some more nice stuff like printing of subject, sender... 
See tracker #868714 "added stats option to archivemail", which has a point.
Message-Ids are useful for diagnosis, but not very nice to read for humans. 

Be a nicer citizen with respect to mailbox locking. 

Perhaps prune/shorten IMAP mailbox URLs in messages?
They may be quite long and may contain the sensitive password.
Also shows up in the process list... 
Perhaps find a clean, lean replacement for all that clutter in the IMAP urls.

Require --output-dir for IMAP archiving?  Otherwise we just drop the archive in
in the current working directory.

Check all items below, which are from the original author. :-)

.archivemailrc support

When you get a file-not-found in the 6th mailbox of 10, it aborts the whole
run. Better to fail gracefully and keep going.

Add more tests (see top of test_archivemail)

We need some better checking to see if we are really looking at a valid
mbox-format mailbox.

Add an option to not cut threads.

Add MMDF mailbox support

Add Babyl mailbox support

Add option to archive depending on mailbox size threshold 
- is this a good idea?

Add option to archive depending on number of messages
- is this a good idea?