Re: two ideas...

> Note - the server rules imply that cache updates
> arrive on a different IP port than direct requests,
> and that the cache loads come on different IP ports
> than direct responses.
> 
> Using these rules avoids the mistake you observe with the proposal
> in HTTP-NG - using the same port requires a RTT to preempt a connection
> and enable a direct response.


PS - note that this *still* takes a penalty of up to two packet times,
worst case. Even if you have different ports for messages to arrive
on, they still come over the same wire. Unless (as in Ethernet) you
can preempt a packet in transit, there's always the possibility that:

	a cache update packet is in transit to the server before
	your direct request
		transmitting your cache update packet would have
		to be preempted by your transmitting a cache update.

	transmitted speculation packets would have to be 
	preempted at the server as well.

and few protocol implementations provide for preemption

Joe