2.1 - 2.1.pl1
	- minor bugfix in restore script


2.1.pl1 - 2.1.pl2
	- Bugfix in configuration programs.
	  Affects:    cartridge handling systems without a special
	              command to set the cartridge, that therefore
	              must be set to sequential mode.
	  Workaround: Comment out the line starting with Setcart-Command
	              in .../backup/server/lib/backup.conf

2.1.pl2 - 2.1.pl3
	- Fix of stupid bug produced while fast-hacking pl2

2.1 - 2.2
	- automatic tape labeling and consistency check
	- directories can selectively be excluded from backup
	- cleverer device access, two stage timeout: 1st stage
	  sends email to maintainer, 2nd stage exits
	- facilities for recovery from hard crash with lost
	  filename logfiles
	- reduced unnecessary tape movements

2.2 - 2.3
	- Maintain a lock file to prevent concurrent tape access of
	  different server processes
	- Optional compressing of the filename logfiles
	- Client side parameters RootDirectory and BackupHost can
	  be overridden with command line options of restore

2.3 - 2.4
	- Block and character device special files can be saved
	- Storing the file contents can be forced independently of
	  the file type (useful for saving raw partitions or so)
	- option -N for the client side program for saving only
	  files newer than a given one
	- small bug fixed concerning zero-length files

2.4 - 2.5
	- File contents (see 2.3 - 2.4) can be compressed
	- Sockets can be saved
	- full_backup, incr_backup and restore are now binaries
	- print_errors is a new command
	- Weird problem workarounded with I/O-Error on interrupted
	  backup
	- If filename logfiles should be compressed, no uncompressed
	  files ever appear
	- Directory path patterns can be supplied for excluding files
	  or directories
	- Support for a max of 16777215 cartridges with 2147483647
	  files on each (please consider configuring several backup
	  servers ;-) )
	- Restore can be called with glob-style filename patterns
	- For the client side a lockfile is maintained as well, so
	  several backup or restore programs can't be run simultaneuosly
	  on one client causing trouble on the logfiles

2.5 - 2.6
	- File contents can not only be specified with prefixes /../
	  or //../, but also by option -r or -R
	- Saving files can be limited to device/filesystem

2.6 - 2.6.1
	- New client- and serverside parameter EncryptionKeyFile
	  for the authentication mechanism
	- New clientside-parameter VarDirectory for varying files
	- Default-configuration files in /etc/buclient.conf and
	  /etc/buserver.conf

2.6.1 - 2.7
	- New clientside parameter CompressBackupedFiles for choosing
	  compression of files and filename logfiles separately
	- Clientside request-queue for asynchronous communication with
	  the server (hopefully much better performance)
	- New serverside command "label_tape" to manually write a label
	  to the tape actually in the drive
	- Default-configuration files can also be (Debian needs)
	  /etc/afbackup/client.conf and /etc/afbackup/server.conf
	- Bug (typo) fixed: Entry for locking file in the server
	  configuration file has been ignored
	- HOWTO-file is maintained now

2.7 - 2.7pl1
	- Better connection failure handling. Up to now with DAT data
	  on tape might be lost on a connection breakdown

2.7 - 2.8
	- Normal filesystems can be used for backups instead of tapes,
	  thus also removable media
	- A file with a special name can be placed in any directory
	  containing names or patterns of files and subdirectories to
	  be excluded from backup
	- Compression attempts can be avoided on already compressed
	  files by supplying appropriate filename patterns
	- Slight further speedup
	- Small bug fixed (sometimes a "Connection ... lost" occurs)

2.8 - 2.8plX
	- Bugfixes, more logging, more diagnostic messages, more
	  safety

2.8 - 2.8.1
	- Port number can be configured in the configuration file
	  of the client side

2.8.1 - 2.8.2
	- FreeBSD supported
	- Wrapper script for mt added: __mt (FreeBSD's mt does not
	  accept 0 as repeat count)
	- Non-GNU-make can be used (under Solaris it's a little
	  annoying) and is recognized automatically

2.8.2 - 2.8.2.pl1
	- Bug ignoring several -m or similar in file list fixed

2.8.2 - 2.9
	- Several sets of cartridges can be used on the server side
	- Verify is now possible
	- restore does not restore the same file several times
	  from different backups
	- memory leak fixed (a quite harmless one, quasi more a
	  theoretical one, no sole reason for an upgrade)
	- Documentation reviewed and lots added, e.g. an INTRO
	- An exit command can be configured, usable for reporting
	  successes and statistics to an administrator
	- Ordinary users can restore their own files and directories

2.9 - 2.9.pl1
	- With several cartridge sets configured reading over end of
	  tape of the last cartridge change to the first cartridge
	  of the same set and not to cartridge 1. This was a bug

2.9 - 2.10
	- Byte Sums are no longer of type "long int", now "double"
	  ("long long int" is a portability problem)
	- Buffering also on the server side. Especially read
	  operations (verify, restore) should be much faster now
	- Stale lockfiles do not cause another server process or
	  a highlevel client program to wait. This was a bug
	- Several kill-s to the server don't harm, no stale lockfile
	  is left over (of course not valid for a kill -9)
	- Errors are immediately printed before completion of client
	  programs. print_errors is not really necessary anymore,
	  anyway i consider it helpful for later diagnostics
	- No longer confusing message "Cannot find out number of
	  lines in ..."
	- No longer annoying message "/tmp/xxxxxx: No such file
	  or directory"
	- Safer tape handling (hopefully)
	- Strange AIX problem workarounded (EIO at end of media
	  instead of the documented ENXIO)
	- A modified-before- and -after-date may be supplied with
	  the restore program
	- Autodetecting inserted cartridge, if tape is labeled or
	  had already been used
	- Strange wish behaviour (too flat window) workarounded
	- Missing feature added: modification time is saved with
	  file contents i.e. e.g. raw devices

2.10 - 2.11
	- Second possibility to save backups in a filesystem
	  (without the need of symbolic links)
	- Output of XXXXXXconfig is piped through more for
	  convenient configuration from any terminal

2.11 - 2.11.1
	- libc-6 (i.e. glibc) - ready (see README: THANKS TO)

2.11.1 - 2.11.2
	- Insufficiency of the cartis script fixed: read the number
	  of cartridge sets from the server configuration file
	- (x)clientconfig and (x)serverconfig did not look for config
	  files in all intended places, documentation incomplete

2.11.2 - 2.11.3
	- Minor inconsistencies fixed. Utilities added (yet unused)

2.11.3 - 2.11.4
	- Security hole fixed

2.11.4 - 3.0
	- Debianized
	- Defaults changed somewhat
	- Uses autoconfig
	- full_backup, incr_backup, restore and verify can override
	  most of the config file settings by commandline options
	- New parameter: Device-probe interval
	- New parameter: Skip-files Command
	- Documentation reworked (including man-pages)
	- restore now takes -C instead of -R, -R is still recognized.
	  -C is now the changedir-option in conformance with the
	  other commands (and BTW tar)
	- label_tape has new -q option for reading the tape label
	- Emergency restore without any input information (-E)
	- Emergency restore or only the minimum restore info (-El)
	  or only tape scanning (restore -En)
	- Differential backup with option -a for incr_backup
	- Hopefully more speed by use of appropriate TCP-options
	- Option to use 128 Bit DES encryption for client-server
	  authentication (requires Eric Young's libdes)
	- cartready indicator file mechanism changed (other way round)
	- Logging of the remote host name, if authentication fails
	- Possibly conflicting program names prefixed with af
	- Bug fixed setting the wrong cartridge for a set, if one set
	  comprises of cartridge #1 only, what should be avoided anyway
	- Senseless settings of maximum number of bytes per tape file
	  caught and reset to reasonable value
	- Year 2039-ready i.e. any length time_t on any architecture
	- More appropriate filename handling on recursive extraction
	  of directories
	- Preferred serving of repeated requests from the same client
	  to avoid race conditions
	- Better locking algorithm for client start and server access
	- SunOS 4.XYZ supported platform
	- Symbolic links are allowed for all programs (full_backup, ...)
	- On remote start the output of the remote command is transferred
	  to the client side, also the exit status corresponds
	- Generally better diagnostics for remote start
	- Very stubborn problem on HP-UX-9 fixed, maybe the problem was
	  leading to problems also on other architectures
	- New parameter: InitProgram for the client side. StartupInfoProg
	  is now ran later, when backup succeeded
	- In serverside init and exit command %p is replaced with client's
	  peername

3.0 - 3.0.1
	- New flag -w for afclient to check server side streamer state
	- differential backup with incr_backup -a did not work yet (!)
	  (implementation was lost probably due to multiple development
	   locations and inconsistent code between them, sorry)

3.0.1 - 3.0.2
	- New flag -l for afclient to report the location of the server
	  side for each file (preparation for storage pools)
	- docs updated

3.0.2 - 3.0.3
	- restore -f was broken in 3.0.1 - 3.0.2

3.0.3 - 3.0.4
	- Server side rexec path had wrong default

3.0.4 - 3.1
	- Several triples backup-host/backup-port/cartridge-set possible,
	  all handled transparently to the user
	- Server side lock bug fixed (caused minor requests like -q to
	  hang, had to interrupt with Ctrl-C, bug appeared in 3.0.3)
	- Also scripts (e.g. clientconfig) may be arbitrarily symlinked
	- Preinstalled config file has now empty encryption file - caused
	  confusing log message (" ... using compiled-in key")
	- Install script supports change of afbackup service name
	- Docs adapted
