As stated in the ENskip 0.50pa "BUGS" file:

  "Remember: This is the fourth release of ENskip (pre-alpha 0.50), and still
   somewhat shaky."

This is also true for this release, albeit to a lesser degree.

This release has been tested under Linux and Solaris ONLY.

The following features are still missing from this release:
- The SUNSCREEN mode is not integrated and probably never will be
- The raw AH/ESP mode is not supported
- Neither ICMP type 40 ("authentication/encryption needed/failed") messages
  nor SKIP_ICMP messages (for algorithm discovery "ADP") are supported
- Fragment reassembly works under Linux only (the kernel does it).
- Certificates are not verified (use UDH certificates)

The following bugs might have survived my latest tests:
- Misc. bugs
  The following non-fatal messages might appear in the kernel log (Linux):
  "Couldn't get a free page" or
  "Couldn't allocate a sk_buf of size xy" 
  You should reboot, this indicates a memory leak. However, I have not seen
  these messages in the current version. 
- Crashes 
  The following messages indicate fatal bugs (Linux):
  "skpush: under xx:yy" 
  "Aiee: Scheduling in interrupt"
  "release: xxx kernel stack corruption. Aiee"
  I sincerely hope you will never see these. The current version seems to be
  stable, however.  Reboot if you can.
- Kernel statistics (Linux)
  The statistics of incoming/outgoing/forwarded packets is wrong.
- No testing (Linux)
  Some of the features haven't been through thorough testing (MTU discovery).
- Random numbers
  The implementation currently does not use a decent/real random
  number generator (the seed is gettimeofday()). 
  Under Linux, support for the built-in /dev/random, /dev/srandom should 
  be added.
- Detach (Solaris)
  Detaching an interface seems to crash the Solaris kernel (it works under
  Linux).
- tcpdump (Linux)
  tcpdump shows also unencrypted versions of all incoming packets. This is
  caused by netif_rx() which is used to feed the packets back into the input
  queue. Consider this a feature, not a bug.
- SMP Support (Linux)
  This implementation clears the interrupt flag to avoid kernel stack overflows
  and to implement "semaphores". This should be fixed for multiprocessor 
  systems.

// 25 Feb 1997 Robert Muchsel <muchsel@acm.org>
