Re: [v6ops] Updating RFC 7084 - alternate logic

[Alexandre Petrescu <[e-mail]>]

Le 05/12/2022 à 14:45, Gert Doering a écrit :
> Hi,
> 
> On Mon, Dec 05, 2022 at 02:31:21PM +0100, Alexandre Petrescu wrote:
>> I meant there is a WiFi Access Point in router mode between
>> the snac router and the ISP's inhome router.  It is because of that
>> additional intermediary router that DHCPv6-PD allocations of /64s might
>> not work.
>>
>> To facilitate my expression, allow me to draw a picture:
>>
>>                   /-----------\      ----       ------------        -----
>>     Internet --- |ISP's inhome |--- | AP |-----| snac router|------|IPv6 |
>>                   \  router   /      ----       ------------       |light6
>>                    -----------                                     |bulb |
>>                                                                     -----
> 
> If there is an AP in "router mode", that AP needs to do "router things",
> which means "handle DHCPv6 PD from the snac router to the ISP's router".

True.

But it is not that straightforward.

For that to work, one possibility would be a for the AP in router mode 
to run a DHCPv6 Relay.  But as far as I know there are no WiFi access 
points that run DHCPv6 Relay software.

Another possibility would be that the snac router is precofigured with 
the address of the DHCPv6 server (ISP's inhome router), so it does not 
need to discover it across IP hops, because multicast discovery across 
IP hops does not work.  And, once the prefix is obtained for the snac 
router, there would be a need for the AP to dynamically maintain its rt 
table with respect to that prefix, i.e. run a routing protocol.

There would be a need to simultaneously allocate prefixes and maintain 
routes for them, an operation often reserved to skilled human planners, 
I think.

> No different from any other combination of routers on a string - each
> of the intermediates need to receive DHCPv6 PD requests from "downstream"
> routers and forward "upstream" (and on the way back, install routing table
> entries, of course).

Well, I agree that the situation is not different from other similar 
situations of routers in a string in other networks than home networks.

However, cascaded deployments of DHCPv6-PD, like 
Server-Client/Server-Client... with a few Relays sprinkled here and 
there are far from common.  When they work, they are hand tuned to work, 
because there is no dynamic autoconfiguration mechanism for such deployment.

Alex

> 
> Gert Doering
>   

        -- NetMaster