In your previous mail you wrote:
   
   I was hacking a bit on integrating NRL's IPv6/IPsec stuff in with
   NetBSD, when I noticed some oddities. As an example, in in_pcb.c,
   4.4lite (and the NRL stuff) has a bunch of code in in_pcbnotify that
   we don't have in ours.
   
   I did some poking around and I started noticing all sorts of other
   subtle and not so subtle differences.
   
   Is our networking code really off a 4.4lite code base? In either way,
   anyone care to volunteer to answer some questions about why we do
   things a little bit differently?
   
=> in fact in BSD 4.4 OSs there are some different parts in
the network code:
 - in_pcbnotify (with more or less arguments, in fact it is a fix
 in last 4.4BSD then use the version with many arguments)
 - in_pcblookup (there is a split version with 3 (?) functions
 in place of the classical overloaded one)
 - radix code (Keith Sklower has published a special version
 with needed fixes for supernets)
 - TCP (security fix against sequence prediction attack, T-TCP, ...)

I have worked with all the 4.4BSD derived OSs (4.4BSD itself,
NetBSD, BSD/OS, FreeBSD, ...) then I can say these differences
are minor and you can mix the best of all the ideas with problems.
I know the NRL's IPv6/IPsec but it is not available outside USA
and the IPsec stuff is strickly forbidden in France (Argh!) then
I can't give a direct help... Sorry

Regards

Francis.Dupont@inria.fr

PS: if you are volunteer, we can talk about all the small fixes
or improvements we can add to NetBSD network code. I believe
we can fix the "Serious longstanding problem with TCP" if
it is not the traditional RFC 1323 mismatch...