Re: soft state (was Re: shim6 and bit errors in data packet headers

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

Iljitsch van Beijnum wrote:

> The reachability testing protocol needs to have access to the shim  
> state because it may need to discover unidirectional reachability. If  
> we assume A1, A2 and B1, B2 where A1 -> B1 and B2 -> A2 are the only  
> working paths, doing a simple ping or something like it won't find  
> these paths.

We can do that without involving the context as outlined in Jari's 
draft. And we need to do it without the context to handle cases when we 
need to do locator pair testing and the peer doesn't yet have context 
state. (See my response to Marcelo.)

> People may also want to limit reachability testing to known  
> correspondents for security reasons.

We definitely need to be concerned about both state creation DoS, as 
well as using the testing protocol as a reflector.

   Erik