OpenBSD IPv6 problem

[Robert Elz]

    Date:        Wed, 07 Mar 2001 09:38:20 +0900
    From:        itojun@iijlab.net
    Message-ID:  <19755.983925500@coconut.itojun.org>

  | 	the problem we see in "A B 127" (or 64) is ambiguity.

Actually, in that one particular case (127), there should be none.  With
a shorter subnet mask there would be for sure (for anyone who doesn't
see this, imagine 3ffe:9000::8/125 - configured with
	3ffe:9000::9 -> 3ffe:9000::A
and then ask yourself what you do with a packet addressed to 3ffe:9000::B)

But where the mask is 127, there are only two possible addresses, mine
and his - I'm not sending packets out at all if the address is mine
(unless I like to implement that as a connectivity test and loop
them through the peer) - in the other case, the address is his, and
sending it to him (the peer) is clearly correct, whatever reasoning
gets you to send it that way.

Given that people want to be able to configure things this way
(it suits their sense of aesthetics) and that it should do no
real harm, I'd probably allow it - and then just treat it internally
as if the prefixlen were 128 (but keep the /127 to return when
requested so people don't get confused).

I might even want to allow /126 for the same kind of reason, though
that starts to get messier.   Nothing shorter makes sense though.

kre