Re: rfc4941bis: Invalid addresses used by ongoing sessions

[Fernando Gont <[e-mail]>]

On 11/2/20 04:53, Pascal Thubert (pthubert) wrote:
> Hello Fernando;
> 
> There is a basic problem in the network probing if an address is still used by a host. Starting day 2, the host using default setting should start using a new address and conserve the older addresses in case there's an active connection. If the network probes the address to figure if it's still there, it actually revives it, meaning that in terms of LRU the old address becomes fresher for all participants, including the stack.

I see the situation as follows:

Case 1: The address becomes invalid, and is removed by the host

The connection is "torn down" (?) on the local host. A remote host that 
had an ongoing session still tries to use it, until eventually the 
connection times out.

Result: the connection fails, and while the address is probed, the LRU 
timestamps are refreshed.

Case 2: The address becomes invalid, but since there is an ongoing 
session, the address is not removed.

The ongoing session survives, and obviously the LRU timestamps are 
refreshed.


With the current status quo, Case 2 has the benefit the the connections 
is not torn down... and I'm not sure it has many downsides when compared 
to #1.

Of course, I agree that if the host could inform the network that the 
address is gone, then it could prevent the LRU timestamps updates. I 
guess this is what you were referring to?

Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492