Re: [dhcwg] WGLC for draft-ietf-dhc-addr-notification - Respond by September 13, 2023

[Ole Troan <[e-mail]>]

Jen,

>>> This is becoming more like a topic for v6ops, but...My point is that
>>> even if all enterprise devices support DHCPv6 IA_NA (let's say there
>>> are no Android devices on the network), that it's not possible to
>>> build an DHCPv6-only enterprise network. SLAAC is required and will be
>>> required at least for some time. So it's not a matter of supporting
>>> DHCPv6 IA_NA or not.
>> 
>> I don’t think that statement is correct in the general case.
>> For your statement to be true, you would have to assume IPv6 only on the link and you would have to assume 464XLAT.
>> 
>> 464XLAT isn’t the only way to bring IPv4 connectivity to IPv4 only applications.
>> I would imagine you could also build a 464CLAT that can share address with the host address.
> 
> I'd expect such approach to have some performance implications, so I'd
> consider it undesirable.
> Anyway it's not how modern OSes do it.
> 
>> You could even give out two addresses in DHCP IA_NA if that was required.
> 
> In theory yes, in practice I doubt it would be possible any time soon.
> 
>> I think you can run a network with DHCPv6 only.
>> In face I’m sitting on such a network right now. Albeit with IPv4 also enabled.
> 
> As I said in another email on this thread: do we really want to assume
> that IPv4 is always around as a safety net for cases when IPv6 is
> broken?
> Is it a desirable deployment scenario? (spoilers: I strongly believe it isn't).
> IMHO when we develop something for IPv6, we shall make sure it works
> w/o any IPv4 on the link.
> 
> So yes, when I'm saying "it's not possible to build a DHCPv6-only
> enterprise network" I do assume it's IPv6-only. Currently it also
> implies 464xlat.

Going by that logic. I think you could simplify your statement to:
It’s not possible to build an IPv6 only enterprise network. Period.

Not all hosts support 464XLAT. Nor any other IPv4 transition mechanism.
Not all networks support the mechanisms for NAT64 discovery.

In the long and arduous path towards that, if it ever happens. I am not at all convinced that the current protocol under discussion, the DHCPv6 address notification protocol is helping rather than harming.

O.