JINMEI Tatuya wrote:
To address rfc2462bis issue 271, "An implementation may want to use
stable storage for autoconfigured addresses", I've added the following
new (sub)section:

================================================================
5.7 Retaining Configured Addresses for Stability

It is reasoanble that implementations that have stable storage retain

nit: reasonable


   their addresses and the preferred and valid lifetimes if the
   addresses were acquired using the stateless address
   autoconfiguration. Assuming the lifetimes used are reasonable, this
   technique implies that a temporary outage (less than the preferred
   lifetime) will never result in the node loosing its global address
   even if the node were to reboot. This will particularly be useful in
   "zeroconf" environments where nodes are configuring their addresses
   by the stateless address autoconfiguration but all communication is
   closed in a single link. In such a case, the failure of a "router"
   (that provides the prefix for address configuration) is not
   significant, but loosing the global addresses might be a pain.

   When an implementation tries to reuse a retained address after
   rebooting, it MUST run Duplicate Address Detection for the address
   under the criteria described in Section 5.4, as though the address
   were just configured by stateless address autoconfiguration. The
   reason for this is because a different host may have started using
   the address while the rebooting host cannot respond to Duplicate
   Address Detection from the other host.

So the host stores its address on stable storage across reboots? That's fine in general, I think, but I read the original issue slightly differently, quoting from what Erik wrote in the e-mail stored at rt.psg.com:

Another thing that can be contemplated relates to the "zeroconf" case
(where some implementations might end up unnecessarily using smaller scope
addresses when the "home" temporarily is disconnected) would be a
recommendation that implementations try to provide address stability using
stable storage. For instance, I think it is quite reasoanble that
implementations that have stable storage retain their addresses and expiry
times (preferred and valid) whether the addresses were acquired using
RFC 2462
or DHCP. Assuming the lifetimes used are reasonable this techniques implies
that a temporary outage (less than the preferred lifetime) will
never result in the node loosing its global address even if the
node were to reboot.

Perhaps there was a discussion about this on the list that I missed, but I think there are two separate issues here:

1) Router and connectivity goes away for a while.
2) Host itself reboots.

Stable storage as such addresses primarily the second scenario. But
I think the proposed text may not be very clear with respect to what
exactly is going to be done to assist the first scenario. For instance,
if you delete your only router from the Default Router list, what is
impact on the global addresses do you wish to have? Deprecate the
addresses, remove them, continue as-is, or what? What if the host
does NOT reboot but the router just goes away for a while, hopefully
we are using the stable storage in this case too, but what about DAD?

--Jari


-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------

Reply via email to