[ppml] Proposed Policy: PI assignments for V6

[David Kessens]

Ron,

On Thu, Dec 09, 2004 at 10:23:39AM -0500, da Silva, Ron wrote:
> > so, i guess that the right thing to do would be to assign a /48 to
> each
> > mobile phone, right?
> > 
> > since RFC 3177 states that:
> >        -  /48 in the general case, except for very large subscribers.
> >        -  /64 when it is known that one and only one subnet is needed
> by
> >           design.
> >        -  /128 when it is absolutely known that one and only one
> device
> >           is connecting.
> > 
> > and since  we cannot be sure that a mobile phone will have a single
> > subnetwork behind, i guess that the /48 would be the proper choice
> > 
> > I am sorry but i think this is a huge waste.
> > For the time being, most of mobile phones will not have a manet
> behind,
> > not one not many subnetworks. I think there must be something very
> > wrong if we assign a /48 to each mobile phone
> 
> From a wireline connectivity perspective, the /48 would be assigned to
> the provider gateway (e.g. DSL/Cable router).  Any subnets downstream of
> that connection would be allocated from the /48 pool.  If there was a
> device within the home (say a multi-modal handset with 802.11x, GSM and
> bluetooth) that needed a prefix (or two) for additional subnetting then
> there should be a mechanism to allocate multiple /64's to that device
> from the /48.
> 
> For consistency, can someone extend that concept to a GSM/CDMA
> connectivity model?  Where is the /48 assigned?  Presumably, whatever
> gateway device is being employed for assigning IPv6 addresses to
> handsets would be assigned the /48 and then any further delegation would
> be accomplished with downstream /64 assignments.

Your assumptions are pretty close to what the current thinking is
(from a 10 mile view, I don't think it contributes a lot to go into the
details here).

> Anyone have any URL's that detail current IPv4/IPv6 addressing plans for
> GSM/CDMA architectures?

I don't have anything available. I am actually trying to get materials
on this topic and can post it if I get some (but please be patient -
I haven't found anyhting useful yet).

On Thu, Dec 09, 2004 at 11:10:24AM -0500, da Silva, Ron wrote:
> > That makes more sense to me, but the question is how is this
> considered
> > when applying for address space on a rir? if a mobile operator
> > claims
> > that he needs a /48 for each handset, does he get the address
> > space
> > requested?
> 
> I don't think I was advocating a /48 for each handset.  What entity
> on
> the mobile network does the address allocation?  That component
> could be
> assigned the /48 and then in turn the handset is assigned a) /128
> from a
> /64 segment on the assigning device b) /64 subnet from the /48 block
> /or
> c) multiple /64's from the /48 block from the assigning device.

b) & c) is what I have heard about, while /48 might be the way to go
if we get real mobile Internet access devices that connect a lot more
than a few devices.

Note that there are also other mobile systems than the 3GPP case, and
that implementations/operational details can vary, so don't take my
words as 'all mobiles will work like this'.

David Kessens
---