RE: [ssm] Re: last call comments on ssm-arch doc

["Rolland Vida" <Rolland.Vida@lip6.fr>]

Hitoshi,

> Let's imagine something like an application of a panel discussion.
> E.g., now, there are two senders (S1 and S2) as the speakers, and then
> we do ((S1,S2),G) join. In this case, does this panel discussion works
> as "a single channel" application or "two channels" application?

We obviously have two channels: (S1, G) and (S2,G). The fact that a receiver
sends an IGMPv3 join containing (INCLUDE(S1,S2),G) does not mean that there
is a unified channel, but that he wants to join two channels in the same
time.

You might have another receiver that sends INCLUDE(S1,G) only, so the
channels have to be separated.

> I've thought these answers would be "a single channel" since both S1
> and S2 work for a same application,

The fact that it is the same application or not has nothing to do with it.
Each source has its own channel.

> So, if each answer is "two channels", is it correct that a channel
> MUST consist of a single source except such failover mechanism?

As Hugh said, the channel ALWAYS corresponds to a single source. If we want
a secondary source to assure a failover mechanism for example, it should
have its own channel. But to assure the transparency of the mechanism for
the end-user, the binding between these two channels should be built in the
application itself, and not at the network layer. At least this is what I
understood from Hugh's message, and this is what seems to be correct for me
too.

Regards,
Rolland Vida

_______________________________________________
Ssm mailing list
Ssm@ietf.org
https://www1.ietf.org/mailman/listinfo/ssm