Le 28/03/2013 19:51, Bob Hinden a écrit :
Alexandru,
On Mar 28, 2013, at 9:37 AM, Alexandru Petrescu
<[email protected]> wrote:
Sorry for the late reply.
Le 19/02/2013 22:08, joel jaeggli a écrit :
On 2/19/13 12:40 PM, Alexandru Petrescu wrote:
I think I may need to actually better expose the problem: how
to form IPv6 addresses for vehicles. (yes we know these already
exist: DHPCv6, PRefix Delegation on cellular, stateless
autoconf, NAT, NPT, 64share).
One of the questions I have in that context what special property
of cars make them need a new method?
Well they're different than Ethernet interfaces. One could have
several Ethernet interfaces in a single car. And, cars have their
globally unique space of identifiers which is not EUI-48.
When one tries to make an IPv6 addressing architecture for vehicles
one goes into planning which could quickly overcome the space of
IPv6.
There are very many hurdles to a simple straightforward IPv6
address planning for vehicles.
1 - At most 2^78 vehicles may exist.
That's a lot of vehicles. It is 302,231,454,903,657,000,000,000 to
be exact.
YEs, that's a lot of potential vehicles; likewise: 2^128 is a lot of
computers potentially connected to IPv6 Internet.
The current world population is currently a little over 7 billion
(7,075,000,000). Assuming a few orders of magnitude of population
growth and ownership of multiple vehicles per person, that still
much much much smaller than 2^78. I don't we need to worry about
handling 2^78 vehicles.
That sounds reasonable. Just to note that VIN addressing space does not
prohibit that many unique vehicles to exist.
Where does that number come from?
This 2^78 comes from the simplest interpretation, total use, and most
straighforward conversion of VIN space of 'digits' to binary space.
(The addresing space VIN - Vehicle Identification Number is fixed-length
17 'digits'. Each of the 8 leading digits can represent 33 distinct
values, the 9th digit 10 values, and so on (interpretations of these
digits are publicly available). It makes up for 33^3 * 33^5 * 10 * 30 *
33^3 x 10^4. Approximately 78bits are necessary to represent this
space; different interpretations of the digits of VIN (how we read the
standards), may lead to different results.)
There may be not enough space in IPv6 addressing architecture
space to uniquely distinguish between all past current and future
vehicles.
Likewise, I don't think we need to deal with past vehicles. We do
need to deal with the active vehicles that are likely to be on the
Internet.
I could agree that common sense and other factors should be taken into
account, if one needs a means to convert VIN space to IPv6 addressing
space. Simply because 78bit is too much of what IPv6 could accommodate.
And although it's obvious to many that common sense should be involved,
there is currently no clear idea (or I haven't seen?) about what IPv6
addresses should one put in vehicles.
Should each vehicle hold a /64? a /48?
Should vehicle manufacturer request a /48 from the RIR to deploy in
vehicles? From the ISP? A PI or PA address?
Or should vehicle manufacturer self-generate ULAs for all its vehicles?
Or all?
If one tries to make a recommendation about this, one needs to
understand various needs from manufacturer, as well as the existing
addressing schemes for vehicles (not necessarily IPv6): how VIN is used,
how IPv4 is used in vehicles, etc.
Alex
Bob
Theoretically, the total number of vehicles possible is given by
interpreting the semantics of VIN (17 "digits", some holding max 33
values, others less). Under optimistic interpretations, a trivial
1-to-1 conversion from VIN "characters" space to bit-space of an
IPv6 address leads to 78bits.
(the structure of VIN is : 1 2 3 4 5 6 7 8 9 10 11 12 13 14
15 16 17 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ | WMI
| VDS | VIS |
+--------+-----------------+-----------------------+ but more
details on the net searching Vehicle Identification Numbers)
It's hard to imagine that the first 78 bits of an IPv6 address
designate one particular vehicle. The IPv6 address structure ofers
something like a maximum of 61bits to designate one particular
subnet. And, in a vehicle there is often more than one subnet.
2 - the prefixes obtained from Registries, or from ISP (which one
should I try first?) may come with a price tag. The more
vehicles, the pricier the allocation.
3 - prefixes which are provider-assigned and/or provider
independent may introduce routing churn in the core of the Internet
- if these prefixes are numerous.
For these reasons, we still look closely at the use of ULAs
(instead of globally routable prefixes) and at how to generate
these ULAs meaningfully.
Alex
Alex
Doug
--------------------------------------------------------------------
IETF IPv6 working group mailing list [email protected] Administrative
Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list [email protected]
Administrative Requests:
https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list [email protected] Administrative
Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
