RE: survivability, rewriting

[Erik Nordmark <Erik.Nordmark@sun.com>]

> Ok, what about replicating data packets and include in them a M6 destiantion
> option that means the ping?

Could do that or other forms of piggybacking (an extension hdr which is
followed by the payload packet) if there is a need for this.

> So, if there are data packets available we can piggyback ping into data
> packets (both forward and reverse path)
> My concern is about the sending host having to wait until the ping reply is
> received.

Why would it have to wait?
It can be optimistic (the same way e.g. neighbor unreachability detection
is optimistic) by sending the data packet and separately sending the
reachability test packet. Whether optimizing that to be one larger packet
instead of two is worth-while is a detail we need to talk about once we
understand the bigger picture choices.

> An additional concern is that i don't think that we could trust ICMP end to
> end, becuase of icmp filtering.

Agreed they might not be reliably delivered.

  Erik