Re: Large EDNS0 buffer sizes and IPv6

[JINMEI Tatuya / 神明達哉 <Jinmei_Tatuya@isc.org>]

At Mon, 04 Aug 2008 19:45:18 +0200,
Florian Weimer <fweimer@bfk.de> wrote:

> How do you make large EDNS0 buffer sizes work with IPv6?  I couldn't
> find anything about that in the RFC or in the draft.
> 
> Unlike IPv4, IPv6 doesn't fragment in the core.  This means that the
> source must perform path MTU discovery and fragment on its own.  This,
> in turn, requires client-specific state, which is generally considered
> a bad thing.
> 
> What am I missing?
> 
> Apparently, this was raised by Itojun at IETF48 back in 2000, but I
> don't see how it was resolved.  Implementations with buffer sizes
> greater than 1400 bytes seem to fragment according to the link MTU
> they are connected to, and do not care about the path MTU.  (I haven't
> got native IPv6, so it's hard for me to tell what's going on.)

My understanding is that the name server is expected to set the
IPV6_USE_MIN_MTU socket option (when available) for its AF_INET6 UDP
sockets.  See RFC3542 for this option.  BIND9 does that; unbound also
seems to do so according to a quick naive grep.

---
JINMEI, Tatuya
Internet Systems Consortium, Inc.

--
to unsubscribe send a message to namedroppers-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/namedroppers/>