Re: I-D Action: draft-mishra-6man-variable-slaac-01.txt

[Mikael Abrahamsson <[e-mail]>]

On Sun, 8 Nov 2020, Ca By wrote:

> I believe running a dhcp server supporting 120M mobile subscribers would be
> a nightmare.  I believe dhcpv6 is a larger stateful database and a not very
> forgiving protocol... and client implementations are ... heterogeneous. ...
> and server failover.. messy
>
> None of that exists today in mobile. I feel the effort and scale are
> daunting, so i am not going to do it.
>
> But, getting a knob in the gateway to set the RA to be 60 (or other)
> instead of 64 would be easy. And, that knob could be constrained to only
> compatible device type (chef’s kisses!)... it can be done

If you already know what /56 the customer has (because you assign it the 
same way as the default bearer /64 is assigned today), you don't need a 
full-on stateful dhcpv6 server in the PDN gateway. You can do it 
statelessly and just respond with the same /56 regardless. You don't need 
a full DHCPv6-PD server, you just need a bare-bones DHCPv6 responder that 
takes information from the local pool database that it already uses to 
keep track of what prefix goes to what bearer.

There is also nothing stopping you from sending two RAs, one for the /64 
and one for /56 and then just 3GPP standardize that this means the /56 is 
routed to you. I'd fully support also having an additional flag in the RA 
saying "if you received this RA, it means I routed the entire PIO to you" 
and you own it.

That flag would be useful also on PPPoE or any kind of point to point 
media (like Ole p2p ethernet proposal).

We've discussed this before. Let's go that route and give people larger 
prefixes instead of trying to carve up that poor /64 into smaller pieces.

-- 
Mikael Abrahamsson    email: [e-mail]