Multicast and IPv4-IPv6 co-existence
Stig Venaas <stig.venaas <at> uninett.no>
2007-09-12 09:26:30 GMT
With the transition/migration to IPv6 we will probably for decades have
nodes and networks that either can be IPv4-only, both IPv4 and IPv6, or
IPv6-only. And of course, combinations of those. Below are some thoughts
regarding multicast and IPv4-IPv6 co-existence.
Looking at end nodes (multicast sources and receivers) the principal
problem is that the source and all the receivers must use the same IP
protocol. If you source multicast and some receivers are IPv4-only and
some IPv6-only, you will need to either send the stream twice, both IPv4
and IPv6, or you need some kind of translation.
Having the source send the content twice might be fine for many single
source applications. It becomes more problematic if receivers send
multicast RTCP reports that you want all the other receivers to receive,
or if you have multi-party applications like conferencing. One thing is
bandwidth, the other is that this is not possible if some of the
participants are IPv4-only and some IPv6-only.
We can hope that in the short run all end-points can do IPv4 multicast
(IPv4-only or dual-stack), and that by the time we get many IPv6-only
there will be few IPv4-only. I am far from sure this will be the case,
but this would simplify things.
Assuming that all nodes participating in a multicast session (all
sources and receivers) can speak the same IP protocol, one can avoid
translation. However, while all the end-points might speak the same,
there will probably be a need for encapsulation/tunneling techniques to
make things work through networks that might only support multicast for
one of the IP protocols. It might make sense to encapsulate multicast
from one IP protocol into multicast in the other, however one could