Network Connects Slowly In Kernel 2.6 Using IPv6

[Pekka Savola]

On Wed, 21 Apr 2004, Jeff Pitman wrote:
> On Wednesday 21 April 2004 21:16, Nils O. Selåsdal wrote:
> > On Wed, 2004-04-21 at 15:00, Jeff Pitman wrote:
> > > be obnoxiously slow, both DNS lookups and simple "telnet
> > > www.google.com 80" tests.
> >
> > With ipv6 enabled, the libc resolver will afaik try to
> > resolve hostnames as ipv6 addresses first. For alot of reasons
> > that might take a long time..
> > [...]
> > So perhaps there should be(and maybe there is..) a way to say
> > try ipv4 first in /etc/resolv.conf ?
> 
> This is interesting.  However, the first hurdle is something deeper than 
> DNS because simple connects using "telnet x.x.x.x 80" takes longer  
> with IPv6 enabled.  Editing /etc/modprobe.conf did the trick for me.

Have you looked with 'tcpdump -s 1500 -vvv -n' how the DNS packets 
look like when you telenet to a host?  Difference with earlier glibc's 
etc.?

Guess 1: addition of getifaddrs() and detection whether v6 s enabled 
or not in glibc may cause some problems, especially if kernel where 
such glibc is used doesn't support it or vice versa.

(Guess 2..M): some other weirdness in glibc/kernel changes.

Guess M+1: you may be trying to connect to sites with broken DNS 
load-balancers, see more and examples at:

http://www.watersprings.org/pub/id/draft-ietf-dnsop-misbehavior-against-aaaa-00.txt

Guess M+2: you may be experiencing some other v6-related issue, see a 
couple of issues at:

http://www.ietf.org/internet-drafts/draft-ietf-v6ops-v6onbydefault-01.txt

-- 
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings