There's a failure mode for instance where client starts with an IPv4-only interface and advertises only IPv4 capabilities but server has only IPv6 groups, so it never sends any MC_ANNOUNCE, nor any other multicast-related frames.
Client then gains an IPv6 address and adds IPv6 capability, but unless client knows it's safe to send a new MC_LIMITS there's not a way to tell server, right?
(Or alternately, if client sends an MC_LIMITS to a server that doesn't support multicast (which the client doesn't know in the current spec if it never saw a multicast-related frame) serve will close the connection from seeing an unknown frame.)
One answer would be to add a server-to-client transport parameter that declares multicast support. (Any useful data to include in it? Capabilities flags saying what kinds of capabilities it supports, maybe?) I'm not sure if there's other good options.
There's a failure mode for instance where client starts with an IPv4-only interface and advertises only IPv4 capabilities but server has only IPv6 groups, so it never sends any MC_ANNOUNCE, nor any other multicast-related frames.
Client then gains an IPv6 address and adds IPv6 capability, but unless client knows it's safe to send a new MC_LIMITS there's not a way to tell server, right?
(Or alternately, if client sends an MC_LIMITS to a server that doesn't support multicast (which the client doesn't know in the current spec if it never saw a multicast-related frame) serve will close the connection from seeing an unknown frame.)
One answer would be to add a server-to-client transport parameter that declares multicast support. (Any useful data to include in it? Capabilities flags saying what kinds of capabilities it supports, maybe?) I'm not sure if there's other good options.