prereq's
--------
must have root@mail1's key in newman:.ssh/authorized_keys

converting an account
---------------------
make sure user has no mail waiting to be delivered on cliff
stop postfix on cliff
make sure user has no mail waiting to be delivered on newman
# no need to stop postfix on newman...we won't be getting any new mail anyway
stop cyrus on newman

convert seen and sub files to flat
copy seen and sub files and user spool directory to mail1
run cyrus2courier
move new maildir into proper place
copy sieve scripts into pysieved in new maildir
fix sieve symlink (defaultbc should be renamed to maildir/.dovecot.sievec i think)
update database to point user to mail1

start cyrus on newman
start postfix on cliff

mass conversion
---------------
stop postfix on cliff
any way to flush what's waiting for domain.org without accepting new smtp?
stop cyrus on newman

get username # from stdin, text file, etc.
check for tmp/$username
  if exists, might already be converting this user (do we care that much?)
run middle block above in 'converting an account'
  save output to tmp/$username
touch tmp/$username # to grab exact time it exited
if anything in user conversion exits non-0
  mv tmp/$username error/$username
else
  mv tmp/$username ok/$username
repeat for remaining users

start cyrus on newman
start postfix on cliff

dealing with problems
---------------------
when running cyrus2courier, exit status can be used to detect trouble. i think when mass
converting it would be good to keep the output of all runs in two directories based on
exit status.

  ok/     # runs that exited 0
  error/  # runs that exited non-0
  tmp/    # runs in progress

the files in these directories would be named after the username we were/are converting.
the script(s) should know to check for tmp/$username (and bail if it exists?) and
rm -f ok/$username and error/$username when starting a run to make sure there's no stale
info.