I assume that whatever MARID cooks up _must_ support IPv6. OK?
Of course it would, this is IETF after all :)
(I say this because right now, an SPF record such as "spfv1
ip6:2000:40a:1b/48 ip4:221.214.40.0/25 -all" seems to be a noop. Parsing
stops at an ip6 directive, so -all is never reached.)
SPF made controversial decision to use ":" as separator before specifying
ip address, however in reality you can write libraries to take that into
account and still work with ipv6 addresses just fine, especially since
correct ipv6 address would usually result in the ending of "::" (or /64
before ethernet part) for single ip or "/mask" for range and that can
be used as indicator of the end
With IP6 we can also very easily reach the 512 packet limit for DNS over
UDP.
With IPv6 DNS UDP packets do not leave 512 bytes for data anyway.
And DNSSec does not use UDP and switches to TCP.
But neither one is our concern and is something for other DNS work groups
to look as to if reliance on 512k size of dns udp packets is real problem
and limiting factor that needs to be addressed.
And proposals have been made to increase the size of one packet on the net
itself to 9k as this leads to more efficent use of 10Gb network capacity
(which commercial networks will begin to use soon). For reference see this
info about Internet2 and for more material do search on "Jumbo Frames":
http://www.abilene.iu.edu/rrsum-almes-mtu.html
http://www.abilene.iu.edu/JumboMTU.html
--
William Leibzon
Elan Networks
william(_at_)elan(_dot_)net