Hitoshi Asaeda | 2 May 2005 13:47
Picon

Re: IGMPv3/MLDv2 - How to merge pending state change reports in such cases?

> Case 1:
> STATE0 - INCLUDE(A)
> STATE1 - INCLUDE(B)
> STATE2 - EXCLUDE(C)
> 
> The question is how to deal with the sources in the pending report in 
> order to confirm successive state changes will not break the robustness of 
> the protocol? Does these sources are merely discarded or after the Filter 
> Mode Change report is transmitted ,then schedule a additional 
> source-change report? If so, how to compute the sources in ALLOW, BLOCK 
> record?

ALLOW and BLOCK can be merged in a signle message, so the messages for
STATE0 and STATE1 can be put in a single message if the retransmission
interval is not passed.
But TO_EX cannot be merged with ALLOW or BLOCK message if their group
address is same.
Hence in your case, all pending reports (having the same group
address) are forcely sent out just before the transmission of the new
TO_EX report message.
Note that I'm talking about my implementation (including KAME) here.
Discarding the pending report is incorrect, but how the pending report
is transmitted in that case depends on the implementation I think.

> Case 2:
> 
> STATE0 - INCLUDE(A)
> STATE1 - EXCLUDE(B)
> STATE2 - INCLUDE(C)

This case is same of above. (i.e. forcely retransmit TO_EX before
TO_IN).
--
Hitoshi Asaeda

Gmane