Re: Ephemeral addressing [was Re: 64share v2]

[Fernando Gont <[e-mail]>]

Hi, Ole,

On 11/11/20 09:03, [e-mail] wrote:
> Philip,
> 
> Let me rename this thread as this opens a much larger issue.
> While being able to rapidly reconfigure an end-user network using the layer3 primitives in 6man, I don't think it's worth solving unless also the upper layer problems are solved.

Well, that's exactly why we have layers: to be able to slice the 
different problems, and keep them tractable, among other things.

I guess that if all areas involved would argue "we won't do it until X 
addresses Y", we'd end up with circular dependencies -- and we don't 
need to (and shouldn't).


> - How do TCP connections survive a renumbering event?

They don't, unless MPTCP.


> - How do applications get notified to reconnect?

An API is needed for that. However, in the case where you connect to a 
different network and addresses from the previous network are 
removed/cleaned up, the addresses would be removed, and you'd get an 
error on the corresponding sockets. Many implementations do this, IIRC.


> - Do all applications have to change as a result? Or use a different transport layer?

With the behavior described in the previous bullet, they need not. If 
you want to have something more elaborate, they might -- with possible 
changes ranging from setting/employing a new socket options, to using a 
brand new API.


> - What happens with DNS configuration? Are you assuming everyone has DNS-SD deployed and working?

That's a different issue. What we have for that is dynamic DNS.


> - What about other configuration? Static addresses for example?

If you mean "manually-configured addresses", that's out of scope: by 
definition, those are hard-coded.


> Good luck I say...
> Until then I suggest that we continue (to pretend) that addresses must be long-lived.

While he haven't been trying to tackle the broader issue of what you 
call "ephemeral addressing" (*), I believe both v6ops and even this 
groups itself (6man) have already deviated from that (e.g., 
draft-ietf-v6ops-slaac-renum).

(*) There's not such a thing as permanent address, unless you can 
somehow guarantee that
  1) the node does not/will not move
  2) the node will not add/remove network interfaces (which leads to #1 )
  3) the network topology will not change (when that happens for the 
leaf network, the prefix -- and hence the address -- will change


I don't think #1-#3 could ever be guaranteed. But I'd say that in the 
modern world, you can't guarantee any of them.


       Alice: - "How long is forever?"
       White Rabbit: - "Sometimes, just one second"

           - - Lewis Carrol, "Alice in Wonderland"

Thanks!

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