What about /etc/nsswitch.conf? Do your hosts and ipnodes lines match and
include "dns"?

Than you very much. This solved the problem. I had

        hosts   files dns
        ipnodes files

in nsswitch.conf. Setting ipnodes to 'files dns' helped. I don't know what 
exactly ipnodes do. (The man page says 'same as hosts'.) It seems that 
resolving hostnames to IPv6 addresses is somehow related to 'ipnodes'. 
(Resolving to IPv4 worked just fine for quite a long time with the old ipnodes 
value.)

What is /etc/gai.conf? It looks like address selection configuration for
some other OS. See ipaddrsel(1M) for how to customize IPv6 default
address selection on Solaris (but I don't think that's at all related to
your problem).

AFAIK, /etc/gai.conf is read by the getaddrinfo() function on Linux. 
(http://linux.die.net/man/5/gai.conf) Many POSIX compliant systems provide 
getaddrinfo(), so I thought the configuration file could be valid on 
OpenSolaris as well. I just didn't notice it is called ipaddrsel.conf instead 
of gai.conf. :-) The syntax is a bit different, but it seems to be the same 
stuff.

/etc/gai.conf is mostly absent on Linux, so that getaddrinfo() uses the default 
settings. I use gai.conf to tweak 6to4 address selection. On Linux, 
getaddrinfo() prefers native IPv4 addresses over 6to4 addresses by default on 
machines that have a public IPv4 address. That sounds reasonable, but testing 
your 6to4 router becomes rather difficult. (In some cases, you don't know 
whether IPv4 is selected on purpose or due to 6to4 tunnel malfunction.) That's 
why I override the default settings (by commenting out one line) and instruct 
getaddrinfo() to prefer IPv6 all the time.

This seems unnecessary on OpenSolaris. If I understand it well, the default 
ipaddrsel.conf will prefer native IPv6 over 6to4 and 6to4 over IPv4. Perhaps 
the latter is not the ideal choice (as far as latency is concerned), but it's 
exactly what I want. :-)

Andrej

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to