Re: broken IPv6 code

[itojun@iijlab.net]

>> I similar issue is if you run a daemon and another program is already
>> listening for incoming connections on that port (whether IPv4 or IPv6
>> or whatever), then both daemons will happily run, but only one will
>> accept incoming connections.
>I don't understand. 
>Are you talking about IPv6 and IPv6 case, or IPv4 and IPv4 case?

	I don't have enough context here (yoshifuji cc'ed to us and i have no
	previous emails), but anyway...

	specification (RFC2553) is not clear enough about how AF_INET and
	AF_INET6 sockets should interact, when they try to bind(2) to the same
	port number.  some of kame guys have tried to convince other vendors
	to clarify it, during discussions toward
	draft-ietf-ipngwg-rfc2553bis-01.txt.  however, we could not convince
	other vendors and the specification is kept vague.  it's too bad.

>> IPv6 should be treated as a unique protocol, just like any other
>> protocol supported by the sockets API. If this is not possible for
>> reasons I don't understand, then the API should be changed so that
>> applications can be protocol independent (eg. getaddrinfo shouldn't
>> return IPv4 addresses in this case).

	from standard point of view, the content of the above sentence is not
	documented, and it is up to implementers (so userland programmers
	need to expect different behaviors between different kernels).
	you may want to check BIND9 doc/misc/ipv6 and KAME "implementation"
	document to understand the issue deeper.

itojun