Re: IGMPv3 for really dumb hosts

On Jun 29, 2006, at 7:44 PM, Hitoshi Asaeda wrote:

> The -lite of igmpv3/mldv2 would contribute to the SSM deployment on
> some level. And I could agree to eliminate an EXCLUDE mode (although
> EX(null) is needed).

Let's be clear.  Its elimination of EXCLUDE-filter-mode on the router.
And the reason for this is that it isn't quite clear how to translate
that state into PIM messages to send up the share-path and shortest-path
without causing disruption in the forwarding trees.

The lite spec is really just formalizing the current behavior.
ie, routers have not implemented the exclude-filter-mode.

As far as using just the ALLOW/BLOCK record types, the routers don't
actually gain much considering that (I believe) most have already
implemented an ad hoc v3-lite.

But, from a router perspective, hosts that used only ALLOW/BLOCK
record types would not impose much of an additional burden.

So, the gain would be mostly from the host side, on hosts that
haven't already implemented the complete IGMPv3.  I do not have
enough information to know if enough hosts might benefit.
I'll defer to your experience.

I'm happy either way.

> However, -lite must keep an interoperability with the full versions
> without requiring any change on the full versions.
> If we define different message types or different usage of the message
> types used by the full versions, it may be difficult to fulfill the
> condition.
>
> Actually, I did a quick review of Hui's draft, but unfortunately I
> could not agree several interpretation, because it seems the draft
> proposes different protocol bahaviors and different usage of the
> record types, which implies something new protocol and includes
> difficulties for the interoperability with the full version.
>
> For your question, igmpv3/mldv2 for dumb hosts; as a host-side
> protocol implementer, I'd try to illustrate how -lite version should
> be designed on them.
> --
> Hitoshi Asaeda