* README: Document how to test that the ssh setup to a remote host is
  working.

* backup-repeatedly: Don't sleep between backups.  Print number of
  cycles completed.  Don't retry backups with warnings, as that takes
  too much time.

* fetch-backup-work: Back up kilauea:/boot.

* fetch-backup-work: Back up kilauea, which contains the datorhandbok
  site.

File exclusion on root filesystems didn't work.

* backup-one-task: Don't produce a garbage exclude file when a
  .lysrdiff-exclude file is present on the / file system.

Use our own rdiff-backup program.

* backup-one-task: Use /opt/LYSrdiff/bin/rdiff-backup to perform
  the backup.

* fetch-backup-work: Back up parts of lenin.

* backup-repeatedly: Test for stop and finish flags more often.

* lysrdiff-status: Do it.

* backup-all: Print job number and total number of jobs when stopping
  or holding.

any changes.  Better handling of changed origins.

* distribute-tasks
  (JobInfo.__init__): Initialize __active.
  (JobInfo.set_active): New method.
  (JobInfo.active): New method.
  (JobInfo.destination): New method.
  (tasks_per_destination): New global variable.
  (interactive): New global variable.
  (read_tasks): Update tasks_per_destination.
  (read_new_tasks): Use JobInfo.source() to simplify code.  Update
    tasks_per_destination.  Set lysrdiffpart of info if the source is
    already backed up, to avoid moving a backed up directory to a
    different partition.  Report a fatal error if the destination
    changes.
  (write_task_lists): Only skip jobs if running interactively.  Only
    reorder tasks if running non-interactively.  Have the user confirm
    changes if running interactively.
  (main): New argument: -i (--interactive).  Only read new tasks if
    running interactively.

Use syslog.  Keep track of the origin of backups.  Removed
workaround for manhattan.  Detect crashed servers faster.

* backup-one-task: Removed special case for manhattan.  Store the
  origin of a backup in the origin file, and refuse to back up from a
  different origin (but tell the administrator how to proceed).  Use
  ServerAliveInterval to speed up detection of a crashed server.  Log
  a message to syslog when a backup completes.

* backup-repeatedly: New file.

* lysrdiff-move-obsolete: New script.  Move backups that are no
  longer mentioned in the task file to a separate place.

Guard against running two fetch-backup-work instances at the same
time, and have the operator confirm any changes it finds.
* fetch-backup-work: Atomically replace the list of backup tasks.
  Ask for confirmation before doing so.

Don't backup snes9x-rediffbackup.
* fetch-backup-work: Don't back up snes9x-rediffbackup, which
  seems to be an old backup in itself.

Back up ftp-pub and medreg.
* fetch-backup-work: Back up and / ftp-pub from hanna, and an
  sqldump of medreg from wrath.
* backup-one-task: wrath uses rdiff-backup stored in /usr/bin.

Added support for the --new option, which only runs backup jobs
that have never completed.
* backup-all: New option: --new.
* backup-one-task: New option: --new.

* fetch-backup-work: Back up / and /var of poseidon, and all the
  lyssvn repositories of poseidon.

Better handling of warnings.

* backup-all: New option: --retry.  Only run df on the specified
  partitions.  Added handling of a "hold" file.

* backup-one-task: New option: --retry.  Added -warn, -nowarn and
  -neverwarnfree status files.

* lysrdiff-status: Print number of warnings per partition, and list
  all warnings.

Support starting two "backup-all" at the same time.

* distribute-tasks
  (write_task_lists): New argument: partitions.  Only output data for
    the specified partitions.  Create a lock directory to ensure that
    two distribute-tasks processes cannot clobber the same file.
  (main): Expect partition number on the command line.

Minor.

* Makefile (install-one-task): New target, extracted from install.

Progress indication.  Lock files.  More output.  Run ssh in batch
mode.  Support --failed.  Support a stop file.

* backup-one-task: New argument: msg.  New optional argument --failed.
  Default remoterdiff to /opt/LYSrdiff/bin/rdiff-backup, and add
  special cases for sedesopim, home and uhumhummy.  Remove special
  case for stalingrad.  Handle exclusion files.  Create
  rdiff-backup.out and backup-summary.txt.  Lock the result directory.
  Unset SSH_ASKPASS and SSH_AUTH_SOCK.  Run ssh in batch mode, to
  avoid password prompts.  Print status message including ETA while a
  backup runs.  Use the --force argument to rdiff-backup to make it
  easier to recover from an initial failure.

* backup-all: New optional argument --failed.  Count the number of
  jobs to run, and pass a running total as the msg argument of
  backup-one-task, to get a nice progress indication.  Stop cleanly if
  the file /nobackup/backup.lysator/etc/stop is found.

* fetch-backup-work: Back up sedesopim.

* distribute-tasks (read_new_tasks): Added missing global declaration.

* Makefile: New file
  (BINDIR): New constant.
  (all): New target.  Do nothing.
  (install): New target.

* lysrdiff-status: New file.  Print some status information about the
  backup system.

* distribute-tasks (read_new_tasks): Randomize the new tasks before
  inserting them first.

* distribute-tasks (read_new_tasks): Insert new tasks first, on the
  assumption that it is more important to back up new areas than to
  refresh old areas.

* distribute-tasks (ordered_tasks): New global variable.
  (read_tasks): Store the job in it as well.
  (read_new_tasks): Ditto.
  (write_task_lists): Retain roughly the old order, but move a few
    random jobs to the front of the job queue.

LYSrdiff can now backup up to separate partitions, creates more state
timestamp files, randomizes the backup order, and backs up most user
data.

* fetch-backup-work: Fetch all lsvn, lyscvs, lyswww and home tasks,
  not just a select few.  Fetch roxen tasks.

* backup-all: New arguments: a list of lysrdiffpart partitions to run.
  Check that the correct filesystems are mounted before doing
  anything.

* backup-one-task: New argument: lysrdiffpart.  That argument
  specifies which lysrdiff partition that the backup should be stored
  on.  It is a small integer.  Adjusted the paths to handle it.  Skip
  manhattan, since it is (temporarily) dead.  Store state information
  in a separate directory, and touch flag files for when a backup
  starts and ends, fails and is being attempted.

* distribute-tasks: New script.  This script is responsible for
  assigning new tasks to lysrdiff partitions.  It also randomizes the
  order of the tasks, to make the backup more fair.

Initial checkin of the proof-of-concept scripts used to backup
some of the most important data on Lysator to sedesopim.

* fetch-backup-work: New script.  Define all backup tasks by quering
  auto_home on nema, and listing certain directories on stalingrad and
  uhumhummy.  As set up, this will find home directories, user www
  directories that use Apache (not Roxen), and CVS and SVN projects on
  the stalingrad server.  Note, however, that only a select few of the
  tasks are actually selected, due to the current disc shortage on
  sedesopim.

* backup-one-task: New script.  Run rdiff-backup to back up one task.

* backup-all: New script.  This script repeatedly calls
  backup-one-task for all defined tasks.