Re: [v6ops] v6ops-host-addr-availability: A Little Pushback

[Alexandre Petrescu <[e-mail]>]

Le 23/09/2015 16:03, STARK, BARBARA H a écrit :
[...]

>>>> From a wireless perspective, I don't really see this as being
>>>> what would convince a wireless operator to deploy PD. The case
>>>>  I try to make is that
>>
>> IPv6 in 3GPP environments allocate /64 per UE today.
>
> Really? I'd heard that was true of some deployments. I hadn't heard
> that it was true of all.

It depends how much detail is mentioned.

Yes, in IPv6 cellular networks the UE sees a /64 in an RA - the Prefix
Information Option.

However, that means little about a /64 being 'allocated' to a User
Equipment (UE).  Yes, the network considers that /64 be for one
particular UE; yes, the network uses internally DHCPv6-PD to acquire
that /64.  The first-hop router requests it and then puts it in an RA
towards UE.

But no, the UE does not request it, typical smartphones do not run
DHCPv6-PD and the network does not respond to prototype UEs making
DHCPv6-PD requests.

It can also be considered that a /64 so 'allocated' to a cellular UE
allows UE to configure each single address in that prefix (as opposed to
a laptop seeing a PIO in RA on WiFi, which must DAD it first and which
may collide to other laptops hearing the same /64).  But this is so - UE
has right to each address in a /64 -  because the cellular link is still
a point-to-point link, contrary to shared WiFi - there's no one else on
that UE link.

Another detail: cellular networks running DHCPv6 to UE, along with RA,
allocate an _address_ to an UE.  That address has no prefix length
because DHCPv6-Address does not communicate prefix lengths.  In some
deployment, the first 64bits of the address allocated by DHCPv6 are
different than the PIO in the RA received at the same time.  The UE ends
up with two global addresses: one can be said to _have_ a prefixlen, the
other no.

In this setting, what one needs is DHCPv6-PrefixDelegation for UE, and
the prefixlength must be < 64.  The reasons are: (1) it allows to
tether/route cleanly more than 1 subnet (2) is future proof for 5G links
which will be presumably more shared than ptp, (3) empowers the UE (it
_requests_ a prefix first, rather than being _advertised_ a prefix).

> If it is true of all, then I'm very curious why Apple and Google feel
> the need for this draft. BTW, the link you provided was describing a
> 3GPP standard. The idea that everything described in the standard is
> implemented and deployed in all 3GPP wireless networks that have
> enabled some degree of IPv6 support is not true. There's quite a big
> difference between what gets written in 3GPP standards and what ends
> up being deployed and enabled.

I agree.

>> I've been told that the SP WiFi architecture best practice is to
>> also allocate dedicated per-UE prefix on converged WiFi as well.
>
> Interesting. I'll have to try that out next time I'm in range of a
> attwifi hotspot.

Or maybe try Monzoon-equipped hotels, as an earlier poster pointed to.

Alex