BUGS of pgl:

Minor bugs may be listed in docs/TODO.

- pgld crashes with ENOBUFS recv error under heavy load on slow machines,
  especially after updating the blocklists.
  There are some workarounds to prevent this, or at least reduce the frequency
  of these crashes. Use a combination of them:
  1. Increase buffers:
     Set NFQUEUE_MAXLEN to a higher value, e.g. 4096 or 8192
  2. Increase pgld's priority:
     Set the variable NICE to a lower value, e.g. -10 (max. -20)
  3. Increase the default receive/send window:
     Add this to your system's config (you may use /etc/pgl/insert.sh
     if you don't know a better file)
       sysctl -w net.core.rmem_default=8388608
       sysctl -w net.core.wmem_default=8388608

- pglgui shows the number of IPs and IP ranges that it is blocking. Due to the
  technical implementation this number keeps unchanged if you whitelist IPs,
  although you should think that it gets smaller. So this is just a "visibility"
  problem, but doesn't mean that anything doesn't work.
  Technical explanation: the number of IPs that get blocked relates to the
  blocklist that pgld uses. But the whitelisting is done with iptables rules
  (target RETURN) that prevent that the whitelisted traffic ever reaches pgld.
  --
  2011-08-11, jre <jre-phoenix@users.sourceforge.net>


- "update" doesn't work with OpenDNS
  see e.g. here:
  http://forums.phoenixlabs.org/showpost.php?p=127410&postcount=7
  http://forums.phoenixlabs.org/showthread.php?t=18045
  http://ubuntuforums.org/showthread.php?t=1189320


- pglcmd only monitors if pgld starts successfully until pgld forks
  (daemonizes). Everything that happens after the fork doesn't get catched.
  - Therefore move as much as possible of the pgld start code before the fork.
    Note that e.g. the pidfile generation has to be done afterwards, because a
    forked process has a new pid. [ DONE ]
  - A possible solution would be for pglcmd to wait some seconds, and then check
    again, but this makes starting slow, so this is not applied.
  - pglcmd.wd partly solves this by doing a restart and notifying the user per
    email and in the logfile.
  --
  2009-11-15, jre <jre-phoenix@users.sourceforge.net>
