In message <[email protected]>, Philip Homburg writes: > In your letter dated Wed, 20 Jul 2011 17:35:31 -0400 you wrote: > >I am not sure the specs insist that an IPv6 implementation > >must treat an ICMPv6 Packet-Too-Big for less than 1280 bytes > >as "unrecoverable". (I haven't re-read the IPv6 specs recently.) > > Some services, like big DNS server cannot afford to do PMTU. It requires them > to store the original DNS reply just in case an ICMP comes back.
This was originally noted in draft-ietf-ipngwg-bsd-frag-00 (January 1998) and became IPV6_USE_MIN_MTU (RFC 3542, May 2003). Now if we can just get it into POSIX .... And, yes, nameservers do set this option. FreeBSD's implementation however is partially broken. See: http://www.freebsd.org/cgi/query-pr.cgi?pr=158307. > Without that, the host will see that its first request fails and has to retry > . > Over time, this will happen again and again because big servers cannot > maintain all PMTU state forever. It takes multiple requests usually and yes the state does get lost within minutes on busy nameservers. What should have been specified is that hosts need to fragment packets into roughly equal sizes fragments. This reduces the probabilty of having to deal with multiple PTB packets. Mark -- Mark Andrews, ISC 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: [email protected] -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
