SIP again

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Christian;

> > The presence of IPv4 options in a packet header cause the IPv4
> > packet to be kicked out of the silicon-path of many of today's
> > high speed routers. Given that history, I could easily forsee
> > a silicon-forwarder for IPv6 that kicks packets to the slow
> > path if the next-header field is _not_ one of several known
> > values (on the theory that if the next header is not X, Y, or Z
> > then there _might_ be something in there that the router needs
> > to look at...). This has undesired effects on throughput.
> 
> The IPv6 architecture is pretty clear about that one: the only packets
> that a router is expected to process are those in which the first
> payload is a "per hop option", or those in which the destination address
> is one of the router's own addresses. All other payloads are supposedly
> end-to-end.

That is a correct argument, but is not enough.

The proper conclusion is that none of the IPv6 options are useful.

Per hop options are, undoubtedly, evil.

Others are end-to-end specific to each form of communication,
which means they belong to not IP but transport layer.

							Masataka Ohta