Re: failure detection

[Paul Jakma <paul@clubi.ie>]

On Wed, 17 Aug 2005, Paul Jakma wrote:

> Or define a very simple link^Wpath-state control protocol for shim6.
>
> You could use it to monitor path-state to locators you're interested in.

And yes, this wouldn't be as cool and as much fun as thrashing out a 
whole new set of signal between data-layer and IP and ULP's, but at 
least shim6 would be self-contained and wouldn't need any inside 
knowledge of ULP's and/or hints from ULPs.

And it needn't preclude optimising Shim6 if ULP hints /were/ 
available.

And yes, it involves additional packet overhead, proportional to the 
number of locators you want to monitor. But the overhead scales, at 
worst, linearly to the number of flows you see - in typical operation 
one would hope that some remote shim6 sites would be 'popular' and 
hence overhead amortised over multiple flows - a slight saving of 
overhead.

regards,
--
Paul Jakma	paul@clubi.ie	paul@jakma.org	Key ID: 64A2FF6A
Fortune:
HOW YOU CAN TELL THAT IT'S GOING TO BE A ROTTEN DAY:
	#15 Your pet rock snaps at you.