0.8:
  upgraded to libnet 1.1.1 API.  Add -DOLD_LIBNET to use old code.

0.7.2:
  Cancel the ARP sending thread when we found an answer.
  This was noted by Tollef Fog Heen first as a cosmetic problem, but
  Alexander Zangerl sent me a patch, so I integrated it.  Apparently,
  there was a race condition on some buggy pcmcia adapters/drivers.

0.7.1:
  new and improved edit-netscape-proxy.pl ;)
  It will now edit netscape and links config files, set the netscape
  proxy type to "user specified" and work if there are no netscape or
  links config files.

0.7:
  divine now works with and requires libnet-1.0 (lots of incompatible
  changes in libnet).  Thanks to Chris Kakris for telling me!
  increased tries before divine gives up to 6 (from 3).

0.6:
  source IP is not set to 0.0.0.0
  Matt Behrens pointed out that divines ARP probes might be enough to
    poison ARP tables if the IP is already taken.  Also, they might set
    off arpwatch or similar intrusion detection systems.  I adopted
    Matt's suggestion to use 0.0.0.0 as source IP instead.
  separated ARP probes for own IP from landmark IPs and added a slight
    delay before the landmark IPs.  It is now more probable that the
    host using my IP answers before the landmark IPs.

0.5.3:
  updated BUGS section in man page
  removed warning about using system() from divine.1

0.5.2:
  several suggestions by Jürgen Vollmer
  divine now accepts newlines in /etc/divine.conf
  divine now looks if the IP you would get in that network is already in use.
  However, this only works if the machine that uses your IP address
    answers first.  A real fix would take a greater change to divine, so
    I postpone this until someone submits a patch or I actually witness
    the problem (on my laptop).

0.5.1:
  use exit code 2 when no answers were found.
  Thanks to Stefan Luethje for pointing this out.

0.5:
  changed system() to fork()+execlp() to avoid /bin/sh
  Divine sets PATH=/sbin:/bin:/usr/sbin:/usr/bin to make sure that
  ifconfig and route are found.
  Divine no longer puts the interface in promiscuous mode as this is
  really not necessary and makes Linux put warning messages in the
  system log.
  Documented that divine needs at least Linux 2.1 because of the
  implicit network route.

0.4.2:
  the shell script now gets the description from /etc/divine.conf
  as command line argument.
  Suggested by Chris.  Good idea!

0.4.1:
  fixed comment in divine.conf.dist
  Argh!  I made extra-sure that I added the fields but forgot to update
  the comments in 0.3.  Thanks again to Chris for pointing this out.

0.4:
  ported divine to glibc
  To me this looks like a bug in libnet.  The headers define struct
  ether_addr for libc5 but don't include the file that defines the
  struct on glibc systems.  Anyway, easy fix.  Thanks to Chris Kakris
  for making me fix this ;-)

0.3:
  first public release
  changed to POSIX threads, the timeout is finally working now.
  added external script

0.2:
  test release
  changed to libnet and libpcap, so no more external processes (except
    to set the interface, of course)
  the timeout code did not seem to work, so divine always hung when no
    network is detected

0.1:
  proof of concept release
  used a slightly hacked arp exploit binary to send the arp requests
  used tcpdump to read the arp replies
  on timeout, tried to killall tcpdump and failed for some reason, thus
    leaving tcpdump processes behind all the time. :(