			 RELNOTES of KAME kit

			     KAME Project

		     $Date: 1999/05/01 18:07:09 $


Here is a summary of differences between KAME STABLE 19990131 and STABLE
19990412.

<<IPv4>>
- Removed mbuf leaks and other mbuf-related twists.

<<IPv6>>
- Removed mbuf leaks and other mbuf-related twists.
- Rate limit for icmp6 redirect.
- Reject overlapping fragment on reception.
- More spec-conformant source address selection for NA output.
- MINCLSIZE is decreased to force drivers to give cluster mbuf.
- (FreeBSD 3.1) cisco HDLC support for sppp device.
- (BSDI) merged `goto ours' hack using the routing table from KAME for
  FreeBSD.
- ND6 reachable time is now recomputed in at least two hours even if
  no router advertisement is received.
- Changed to use ND based MTU instead of the link's physical MTU
  when detecting whether the packet should be fragmented.

<<IPv6 API>>
- setsockopt(IPV6_{UNI,MULTI}CAST_HOPS) now behaves as described in spec.
- Fixed libinet6 in order to support a lot of interfaces(e.g. over 100
  gifs can be configured).

<<IPsec>>
- Avoid corrupted packet on tcp retransmission.
- Fixed a bug in IPsec header size computation for MTU consideration.
- esp_output() now takes care of both IPv4 and IPv6 (used to be separate).
- Fix memory leakage in IPsec tunnel decapsulation.
- (NetBSD) tcp4/6 takes care about IPsec header size in MSS computation.

<<Userland>>
- traceroute6 supports source route (-g).
- -c specifies alternate configuration file path for rtadvd.
- Bug fixes and enhancements of bgpd.
- Additions of several manpages.

<<Others>>
- ports/pkgsrc: upgraded base version for many items in the collection.
- ports/pkgsrc additions: wbd, squid, ethereal
- ports/pkgsrc removals: im

KNOWN BUGS/TWISTS
=================
All the following problems will be fixed in near-future SNAP releases,
and the next STABLE release,

- "pim6dd" has not been fully tested yet.  It may have several bugs.
- "faithd" sometimes fail to relay rsh/rlogin connection properly.  Some part
  of data connection seems to be lost.  rlogin session may fail to propagate
  screen size information to the peer.
- Some of the userland tool may not work properly, if you configure more than
  500 interfaces.  (libinet6 is fixed for this, but there are some places
  where max # of interfaces is hardcoded)
- Notebooks/laptops problem: multicast hardware filter on ethernet card will
  not be properly initialized after suspend/resume session, and this makes
  trouble with IPv6 commuincation (which heavily uses multicast).
  This is not a KAME problem (*BSD problem), but please be warned.
  Workaround: perform "ifconfig down", then "ifconfig up" after resume.

							<end of RELNOTES>
