Le 28/03/2013 20:29, Manfredi, Albert E a écrit :
-----Original Message----- From: ipv6-boun...@ietf.org
[mailto:ipv6-boun...@ietf.org] On Behalf Of Alexandru Petrescu
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.
Just out of curiosity, why not use DNS to solve this problem? Isn't
that how these things are normally handled?
Well, it depends how one sees the VIN - Vehicle Identification Number.
I could agree that a VIN could be used to identify applications, or
services. There are current works about converting VIN to a FQDN.
Something like this: the VIN VMP3565AXT3MD4460 is mapped into
gaslevel.T3MD4460.3565AX.VMP.vehiclebrand.fr
Remark the conversion is not straightforward. A selection was performed
on the fields of the VIN; other people may select differently.
You can use the VIN, or you can use the owner's name/address/etc.,
along with an added way of indentifying a system or component, to
create a number of unique identifiers a car might need. It would be
nice if each car had just one, but more than likely, each subsystem
will need its own IP address and DNS record.
Yes, many things are possible, in addition to using VIN. And yes, I
agree each subsystem in the vehicle may need its own IP address and DNS
record. And within each vehicle there are many such subsystems.
Then you use whatever IPv6 prefix makes sense at the time, possibly
dependent on location.
Well at many locations where vehicles could connect to Internet, the
cellular connections are used (less often WiFi and/or satcom).
The cellular connections deliver one address, but not a set of addresses.
Thus one needs a prefix from a 'home' anchoring point.
That doesn't solve the problem - the home still needs to decide which
prefix to tell the vehicle to assign to its computers. What IPv6
prefixes make sense to deploy in vehicles? Should it be PI or
PA? Or ULA? How big the prefix length request should be, and requested
to whom? Is a vehicle a Site? Or is the grouping of vehicles an entire
Site?
Moreover, although that may connect the vehicle to Internet, if each
vehicle did so, then direct V2V communication would be highly
inefficient - even impossible at times (each would have to connect its
home first).
The 64-bit (or whatever length) IIDs can be chosen in any of the
usual optional ways.
Yes, there exist various ways to configure the IID, mostly in
relationship with interface hardware identifiers.
One potential other way would be in relationship with VIN - that's what
this draft is about. (but the VIN could be used to generate ULAs as
well, the other draft).
Like all IPv6, this does not preclude also using ULAs.
Right, we are looking now at using ULAs. If one tries to generate ULAs
by following the RFC, it gives an example in which they suggest using an
EUI-64 in the algorithm. But they dont say _whose_ interface's EUI-64.
A choice could be made among the importance of various PCs in the car,
but then why not choosing the obvious.
But for example, just like OnStar systems are able to do, it will
continue to make sense to allow a stable way to reach each car.
I believe OnStar is the equivalent of eCall in Europe(?)
If so, then this a single computer in the vehicle which has a single
job, and which has a single IPv4 address. If there is other computer in
the vehicle it doesn't connect to the Internet via eCall - it must not.
With IPv6 we look at having several such computers in the vehicle and
each one of them to be reachable from the Internet directly, no
application gateways, but yes secured.
The DNS approach makes it possible to reuse IIDs for cars in
different parts of the world.
Provided that DNS-updates are feasible for large numbers of vehicles
highly mobile, or that 'triangular' long-distance DNS requests can be
tolerated.
Or, we can insist that cars have a single IPv6 interface, and that
all internal components are addressed in a standard way, using
subheaders in the data payload. I just don't think that will fly,
though.
Right, I dont think it would fly either. Advanced IPv4 cars of today
already have many IPv4 addresses inside the car, no need of subheaders
like Source Route. Just that these dont scale to many cars, because
IPv4 not enough addresses left.
Thanks,
Alex
Bert
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
