Subject: Re: IPv6/v4 host to IPv4 host strangeness
To: None <email@example.com>
From: None <firstname.lastname@example.org>
Date: 09/02/1999 14:24:54
>phil (IPv4/v6 host) has -current from the beginning of July (3rd ???) - just
>before the ELF change (I was afraid to break my running machine - but
>maybe indeed it is time now for the change ...).
>search apc.aip.de ipv6.apc.aip.de
>lookup file bind
>default, except the hosts: line
>hosts: files dns
>126.96.36.199 phil phil.apc.aip.de
>(plus some other entries related to other hosts)
>BTW If I add spade.apc.aip.de to the 188.8.131.52 line, the
>telnet spade.apc.aip.de "spade.apc.aip.de: non-recoverable failure
>in name resolution" gose away and behaves exactly like telnet spade.
I believe the "use of bogus IPv6 address" part of problem is already
solved. Please update to more recent code, especially src/lib/libc/net.
Search order is still very tricky part. telnet uses getaddrinfo(3),
and getaddrinfo(3) calls gethostbyname2(3) twice to resolve
"phil" to IPv6 and IPv4 addresses. As a result, search order
IPv6 phil.apc.aip.de -> not exist
IPv6 phil.ipv6.apc.aip.de -> exists
IPv4 phil.apc.aip.de -> exists
IPv4 phil.ipv6.apc.aip.de -> not exist
so, IPv6 wins over IPv4. Actually, getaddrinfo(3) should use
getipnodebyname(3) as its underlying function, however, it is not
merged for avoiding too many non-BIND changes in resolver code.
This was discussed in July 99 in tech-net or somewhere, I believe.