RE: Are IPv6 auto-configured addresses transient?

["Hemant Singh (shemant)" <shemant at cisco.com>]

-----Original Message-----
From: Christopher Morrow [mailto:christopher.morrow at gmail.com] 
Sent: Sunday, October 18, 2009 5:01 PM
To: Hemant Singh (shemant)
Cc: Brian Haberman; Margaret Wasserman; ipv6 at ietf.org
Subject: Re: Are IPv6 auto-configured addresses transient?


>I have a long lived ssh session between two devices, both of which use
>enhanced privacy addresses that rotate hourly.

If the addresses rotate at the hour, then a callback function from the interface can call the app and report the change.  It's not even that an app has to track the interface IP address changes - the app just needs to register a callback to the interface ip address change processing function.  Even in v4 IP addresses can change by the hour and many v4 apps have been developed sensibly.

>(example provided)

>Note that in most cases the ip address stays live on the device as
>long as there's something actively using it... so its not really a
>problem, unless the entire subnet changes (upstream provider change
>for instance).

>> anything related to lifetimes of addresses.  On an IPv6 node, if
>> anything related to the addresses on a network interface changes, the
>> changes will affect the network interfaces the app is using.  Even in
>> IPv4, if an app is using an IP address of a network interface on the
>> node, the app tracks changes of the IP address on the network interface

>The app MAY track this info, it MAY also not :( sadly some people
>hardcode the ip address into the application, of course you can't fix
>that case.

Agreed.  Well, I state a simple coding mechanism above.  The apps have to do better now.  I think for such issues mentioned, some well-known proper coding practices exist. It up to BEHAVE to see what, if anything, needs specification here.  I am still not convinced a problem exists here for which the IETF has to do anything about.  I am definitely convinced on the referral IP address that Brian mentioned in that the referral needs some thought.

Thanks,

Hemant