mailman (1:2.1.9-6) unstable; urgency=medium

  * This version will automatically upgrade indexes of the current
    archiving volume to the new Unicode format, which can lead to a slight
    corruption / data loss in the form of non-ASCII characters not being
    transcoded correctly.

    If you kept single-mbox archives of your mailing list additionally to
    the HTML ones (even if this mbox archive was not downloadable from the
    web), you are probably better off regenerating the archives from
    scratch with:
     cd /var/lib/mailman/lists
     for list in *; do /var/lib/mailman/bin/arch --wipe "${list}" ; done
    This will use information in the original email to make a better guess
    at the encoding of the subject and author of a post to the list.


    Details:

    Some previous versions tried to use Unicode for new entries, mixing
    them with old 8 bit character string entries left over from even older
    versions, leading to a cascade of events that broke archiving for
    some mailing lists completely for the rest of the archiving volume
    period (often the calendar month).

    As the old entries were stored as strings without encoding
    information, there is no way to automatically reliably convert those
    to Unicode. The upgrade code will try, in order:
     - the default Python encoding (usually ASCII)
     - UTF-8
     - windows-1252, a superset of iso-8859-1
    You can change the default Python encoding by creating a file called
    sitecustomize.py in python's loadpath with e.g.:
     import sys
     sys.setdefaultencoding('iso-8859-8')

 -- Lionel Elie Mamane <lmamane@debian.org>  Tue, 27 Feb 2007 21:30:49 +0100

