Re: Announcing Prefix Delegation extensions to ND draft-kaiser-nd-pd-00.txt

[Alexandru Petrescu <[e-mail]>]

Hi,

Le 24/10/2012 02:56, John Mann a écrit :
> Hi,
>
>
> On 23 October 2012 03:54, Alexandru Petrescu
> <[e-mail] <mailto:[e-mail]>> wrote:
>
>     Le 20/10/2012 23:51, Thierry Ernst a écrit :
>
>
>         Dear Alex,
>
>         Would you explain why the vehicle would need to get a new prefix
>         (and
>           thus I assume configure all the nodes in the vehicle) every
>         time it
>           enters a new area ?
>
>
>     Well, whenever MR of a vehicle changes its attachment point it would get
>     a new different address, right?  I can only suppose it would get a
>     different delegated prefix as well.  It's hard to imagine that it would
>     get a different address but a same delegated prefix, no? (it's hard to
>     make same prefix valid at so many different places, harder than doing it
>     with addresses and it's not done with them).
>
>     Or do you ask why LV gets a new prefix when IV changes its prefix?  I
>     think this is obvious, no? (for topological correctness, right?)
>
>     Or do you ask from the NEMO perspective?
>
>     In this V2V2I work we first consider there's no MIP nor NEMO neither on
>     IV nor on LV.  We'll see later about adding MIP.  We can discuss it as
>     well, see how MIP would fit in this.
>
>     Is this answering in the direction you made the question?
>
>     Alex
>
>
> I'm confused about what problems are being solved / created here.
>
> I assume V2V2I is vehicle-to-vehicle-to-Internet.

Yes.

> Why do you _want_ the LFN end devices to change IPv6 address as the
> vehicles move around?

Well, it may indeed be desirable to avoid that change.  But there is 
still a need for initial assignment of IP addresses, no?  Since this is 
IPv6 no NAT, what IP addresses should be initially assigned to devices 
in vehicles?

> How about if you one-off assign prefixes to the in-car subnets, and then
> one-off assign host addresses to the LFNs.

It is a possible way.  When pre-assigning theses addresses one realizes 
though they are topologically correct at only one place in the Internet. 
  I.e. if we assign a vehicle the prefix A which is valid in by lab X 
when I move the vehicle in area Y the prefix A is no longer valid.  I.e. 
one must absolutely use a tunnelling form to connect prefix A in area Y, 
right?  Or, first we don't want to use MIP (we'll see later about MIP).

> Then use tunnels / NEMO / Proxy MIPv6 / whatever to connect the cars to
> the Internet.

Well yes.  But we try it here without tunnels.  This has certain 
advantages: no need to use HA, no tunnelling, no triangular routing, etc.

(I wonder whether there's any other tunneling form than MIP, PMIP, VPN 
and 6to4).

> The LFNs having stable addresses would facilitate connections to and
> from the Internet.

Yes, only if tunnelling is used.

There are additional ways to form stable addresses (we consider VIN, 
more later) in a vehicle.  These can be used to connect vehicles between 
selves forming a local domain, but disconnected from the Internet.  This 
has still advantages over no connection at all.

It would be abnormal to forbid two vehicles in close vicinity to talk to 
each other just because Internet is not available there, or because 
their MIP can not connect to the HA.

For this we use VIN to generate ULA prefixes and 
http://tools.ietf.org/html/draft-petrescu-autoconf-ra-based-routing-02 
to exchange prefixes.

> Is there some soft of association between IVs and LVs?
> - are they owned / managed by the same people?

no, not really.

> - is there guaranteed to always be a IV in range of every LV?

no, not really.  There may be question of several IVs present near one 
LV and vice-versa.

> - are the IV's happy that the LVs are using their bandwidth to the Internet?

Probably yes.  Advertisement billboard vehicles are such kind.

> - is there any need for the LFNs on IVs and LVs to communicate with each
> other? locally?

YEs, like at an incident scene, where law enforcement fire department 
vehicles deployed need to exchange files about a casualty.

> Do e.g. cellular or satellite networks used for connecting IVs to the
> Internet give out different IPv6 address or delegate different prefixes
> as you move around?

I guess yes - different as we move around.

> Or does it take a roam plus a "reboot" to get a new address and prefix?

I guess yes, a roam and a reboot, some times.

Alex

>
> Thanks,
>      John
>
>
>
>         Thierry
>
>
>         On 20/10/12 20:10, Alexandru Petrescu wrote:
>
>             Le 20/10/2012 18:42, Mikael Abrahamsson a écrit :
>
>                 On Sat, 20 Oct 2012, Alexandru Petrescu wrote:
>
>                     One point that guided towards choosing ND over DHCP is
>                     topology. DHCP topology can be relatively complex with
>                     Client/Relay/Server, whereas ND is simpler one-on-one.
>
>
>                 There is nothing saying DHCPv6-PD can't be done in a single
>                 device (the router itself). That's what I do in my home,
>                 cisco
>                 router, local DHCPv6-PD pool, local DHCPv6-PD server, also
>                 installing routes into RIB.
>
>
>             YEs, because at home one typically puts up the interface once a
>             month and gets typically the same prefix from ADSL operator as 1
>             year before.
>
>             But with vehicles, one connects a vehicle here and gets a
>             prefix,
>             then moves in that area and gets another prefix.  At that
>             point, if
>             the router obtaining a prefix wants to delegate further to
>             another
>             vehicle needs to change the delegated prefix.
>
>             This dynamic change between the received prefix and the
>             delegated
>             prefix is not a matter of DHCP.  It can be implemented by like
>             scripting which are independent of DHCP implementation.  One
>             has to
>             touch the conf files be it of DHCP or of ND.
>
>                     _and_ Relay (or Server).  This may be feasible in
>                     practice but
>                     I think it would be cleaner to have distinct
>                     protocols on a
>                     same machine for receiving a prefix and for sending
>                     a prefix.
>
>
>                 What is cleaner is to use existing standards where there
>                 already
>                 is running code.
>
>
>             Right, there is cleanliness in reuse.  Reuse as much as
>             possible.
>
>                     There is also the question of availability of DHCP
>                     software on
>                     smaller platforms which have no SIM card.  It may be
>                     easier to
>                     do this with ND in smaller settings.
>
>
>                 I'd imagine that there already are 2-3 existing FOSS
>                 available
>                 implementations that do what you need for DHCPv6-PD
>                 client and
>                 server. Instead you want to invent a new standard and
>                 create new
>                 code.
>
>
>             In addition to FOSS (what is FOSS?) DHCP one also needs to
>             dynamically change the delegated prefix when the assigned prefix
>             changed.
>
>                 I'm not saying this shouldn't be done, I'm just saying I
>                 don't
>                 really see the rationale for it. I used to hate DHCPv6
>                 role in
>                 IPv6, but after a few years of being exposed to it, I've
>                 come to
>                 accept that this is the way it is. There is code going
>                 back to a
>                 standard Windows Vista that correctly implements DHCPv6-PD
>                 client, and that is what, 5-6 years ago it was released?
>                 I've had
>                 PD in my home on Cisco code for 3-5 years already, with
>                 no server
>                 infrastructure at all, just single device doing
>                 "everything" for
>                 the role needed.
>
>                 If this was 2002, I'd agree with you that ND PD could be
>                 feasable, but I believe the train has already left the
>                 station
>                 and we should focus on keeping IPv6 stable when it comes
>                 to how
>                 it works, and get implementations going, not new standards.
>
>
>             WEll yes, I agree that IPv6 should be kept stable and part
>             of that
>             may be that we try to make sure that a new proposal does not
>             break
>             existing implementation.  This is a matter of further work.
>
>             Alex
>
>             ------------------------------__------------------------------__--------
>
>
>     IETF IPv6 working group mailing list
>
>             [e-mail] <mailto:[e-mail]> Administrative Requests:
>             https://www.ietf.org/mailman/__listinfo/ipv6
>             <https://www.ietf.org/mailman/listinfo/ipv6>
>             ------------------------------__------------------------------__--------
>
>
>
>
>
>         ------------------------------__------------------------------__--------
>         IETF IPv6 working group mailing list [e-mail]
>         <mailto:[e-mail]> Administrative
>         Requests: https://www.ietf.org/mailman/__listinfo/ipv6
>         <https://www.ietf.org/mailman/listinfo/ipv6>
>         ------------------------------__------------------------------__--------
>
>
>
>     ------------------------------__------------------------------__--------
>     IETF IPv6 working group mailing list
>     [e-mail] <mailto:[e-mail]>
>     Administrative Requests:
>     https://www.ietf.org/mailman/__listinfo/ipv6
>     <https://www.ietf.org/mailman/listinfo/ipv6>
>     ------------------------------__------------------------------__--------
>
>