RE: WG next steps

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Mon, 18 Nov 2002, Michel Py wrote:

> Again, what is the issue here: you *might* *potentially* provide free
> transit for the first SYN of Joe Blow surfing a web site that is hosted
> by one of your competitors. Is this kind of bandwidth on your radar
> screen?

To de-confuse the issue, here my understanding of what MHAP does in this
regard (Michel, correct if I'm wrong):

In MHAP, state must be created somewhere close to the source before
packets can be sent to a multihomed destination. The first packet
triggers this. As an optimization, this first packet is tunneled through
the mapping discovery system and delivered to the destination. It's not
much more effort to simply deliver the packet and use it as an implicit
discovery trigger rather than throw it away and create an explicit
trigger packet, but it does help the application a good deal.

If this is too controversial, I'm sure it's possible to do this some
other way. But in that case we may want an ICMP message that goes back
to the source and says "sorry, but we had to use your packet for state
creation and it hasn't been delivered, but now the state is here so try
again".

Iljitsch