Re: [v6ops] I-D Action: draft-ietf-v6ops-ipv6rtr-reqs-00.txt

[Tim Chown <[e-mail]>]

Hi,

> On 8 May 2017, at 03:10, Lorenzo Colitti <[e-mail]> wrote:
> 
> On Sat, May 6, 2017 at 7:01 AM, <[e-mail]> wrote:
>         Title           : Requirements for IPv6 Routers
>         Authors         : Zaid Ali Kahn
>                           John Brzozowski
>                           Russ White
>         Filename        : draft-ietf-v6ops-ipv6rtr-reqs-00.txt
> 
> As Ted has stated many times, this text:
> 
> ======
>  Routers supporting IPv6, and intended for user facing
>    connections, MUST support:
> 
>    o  [RFC3646]: DNS Configuration options for Dynamic Host
>       Configuration Protocol for IPv6 (DHCPv6).
> ======
> 
> needs clarification. Does it mean that:
> 	• A router MUST contain a stateless DHCPv6 server implementation that supports the DNS option?
> 	• A router MUST contain a DHCPv6 relay implementation that supports the DNS option?
> 	• If a router contains an DHCPv6 server or relay implementation, then that implementation supports the DNS option?
> #1 or #2 would be a change to the IPv6 node requirements, which say:
> 
> ====
>    support for DHCP
>    server functionality on routers is optional.  However, routers
>    targeted for deployment within more complex scenarios (as described
>    above) SHOULD support relay agent functionality
> ====
> 
> I don't think it makes sense to say #1 or #2 and thus say that all IPv6 routers MUST support DHCPv6, because:
> 	• Other than for home gateways where DHCPv6 is required, I'd expect that the vast majority of routers don't actually support it at all (at least not in all feature sets).
> 	• In most places in the network there's not even any point in supporting it, because it will never be used. DHCPv6 can only ever be used in routers that serve as first-hop routers for hosts, so a backbone router will never need a DHCPv6 server or relay.
> #3 does make sense.

Worth noting that the node requirements document is undergoing an update, so its text might change.

I would certainly say that #3 is a minimum

There are also of course enterprises that wish to run DHCPv6, but I’d assume they would do so using first-hop relays. In that light, adding at least a SHOULD for RFC6939 would be good, if its not in this draft already. 

The text should probably talk about DHCPv6 requirements for routers that may typically be deployed in client-facing scenarios.

Tim