All things related to multicast

[Tom Pusateri <>]

We don't set the checksum in the dummy ip header.
If the header length of the dummy IP header is set,
then we check the checksum on reception since we think its a real
IP header and not a dummy one. One can argue that this may or may
not be correct but that is how it has always been implemented.

But if the spec is followed as below, the header length will
not be set and we won't check the checksum on the dummy ip header.

>From RFC 2362 as well as the latest internet draft:

        For (S,G) null Registers, the Multicast data packet portion 
        contains only a dummy header with S as the source address, G as
        the destination address, and a data length of zero.

You know its IP version number from the outer IP header.

This does leave the Group/Source exposed since there is no checksum
covering it.

Tom

In message 
<>
 you write:
>>> The TTL of the dummy ip header is not significant. It gets set to 0 when
>>> we zero the dummy ip header and is ignored. The only two fields that
>>> are significant in the dummy ip header are the source and destination
>>> (group) ip address.
>
>    And the first nibble of the header is 4? Right? Because if it is not,
>    you can't be sure the source and destination addresses are where you
>    expect them to be.
>
>    Also, what is the checksum field set to and do you check the checksum?
>    After all, it is a real IP header.
>
>Dino