Re: [v6ops] How do you solve 3GPP issue if neither operator nor handset supports PD?

[Philip Homburg <[e-mail]>]

> Right, so can you know describe how the network, hosts and applications
> are supposed to deal with flash renumbering? Including servers in
> that network.
> 
> Specification, implementation and deployment status?
> 
> I'm drilling down on this to see if we can get a common understanding
> of what the IETF gaps are, and what the implementation and deployment
> gaps are.  (Not making any assumption that the IETF would ever
> recommend an ephemeral addressing model of course.)

Network: 
- DHCPv6 PD: there is a little bit, I think DOCSIS requires the CPE to poll
  the upstream router
- I still plan to read the HNCP spec one day, but I assume it has support for
  with drawing prefixes and annoucing new ones. 
- We can assume that routers get the new prefix from HNCP and use RA to
  renumber hosts.

Host:
- hosts get a new SLAAC prefix using RA, the old SLAAC prefix will be marked
  deprecated
- DHCPv6 has a way to inform hosts, but it doesn't seem widely implemented.

Applications:
- applications have to use short flows and detect broken connections. Many
  applications can do that, but certainly not all
- for applications that use TCP and want to keep the connection, we could use
  MPTCP. 
- QUIC will be seen as a path forward 

Servers:
- by and large servers don't have to anything special. Servers just receive
  requests, it doesn't matter to the server what address is used to reach the
  server.
- For a service, we probably need to update directory information:
  - DNS has an update mechanism. A daemon on the host could detect address
    changes and update. There doesn't seem to be a standard notification
    mechanism at the operating system level. 
  - Torrents use DHTs. This would require OS specific code to detect the 
    change and then the torrent process can update the DHT.