Le 24/10/2012 19:04, STARK, BARBARA H a écrit :
If an LV never ever wanted to get a PD from anything other than an
IV, and an IV could only ever expect to delegate to a LV, then I see
no problem.

I understand in that case there would be no problem use ND instead of
DHCP to realize PD.

On the other hand, if these things do expect the same physical links
 to be used to connect with other ecosystems (like home networks or
hotspots) that will be well-established in the IPv6 world by the time
these vehicles come along, then I think this proposal has some
undesirable ramifications.

Questions: Would an LV ever expect to use the same physical link to
connect to a home network or hotspot, instead of an IV?

YEs, LV would use that egress interface to connect to a fixed hotspot
sometimes, instead of an IV.

Would an IV ever expect to provide connectivity for devices that
think they're inside a home network (e.g., maybe they're in a RV
park, and the RV-owners networks are made up of traditional home
networking components, so they can connect to non-cellular uplinks
when such are available; maybe the IV is the truck pulling a
trailer, and there's a regular home network set up in the trailer).

Well, the trailer is still mobile, a vehicle, I think.

But, even though IV would mostly be like a billboard advertisement
vehicle on a highway whose clients would be other vehicles most of the
time, there may also exist clients other than mobile vehicles.  I can't
see which right away, but why not.

If the answer is no, then read no further. If the answer is yes,
then I'd like to dive a little deeper into understanding the
ramifications of what is proposed.

Hotspots and CE (home network) routers will use IA_PD for prefix
delegation.

Yes.  It would be the case for devices like LTE-WiFi boxes and for CE
customer equipment.

Remark though that for CE, the nature of the uplink - a cable - would
make it hard to see a vehicle on it: it only sees the operator's end.  I
doubt there would be a conflict there.

Sub-delegating routers attaching to them will request prefixes by
IA_PD.

Which would be these sub-delegating routers?  WiFi range extenders?
Others?  I am not aware of which direction these devices take - whether
or not they'd use DHCP to obtain prefixes, or open to something else.

This something else may be open, even to protocols which transform an
IPv4 address into some IPv6 prefix, or 6rd, or so.

Which means either the LV would have to support both ND_PD and
IA_PD, or the hotspot/CE router would have to do so. And whoever does
both would have to know which to use in which case, and make sure
they don't get things confused between the two. Somebody's
complexity-of-code has just been doubled (or more) by the addition of
a 2nd way to do the same thing. If it's the hotspot/CE router, then
it would have to run both delegation mechanisms simultaneously and
keep track of prefixes delegated by IA_PD and ND_PD -- making sure
there's no overlap in delegated prefixes and such.

Well yes, code complexity.  I agree that if there existed two means to
delegate prefixes then complexity is added up.  In that sense, one may
consider that one out of the two ways to realize PD is maybe of lesser
importance than the other, with less ambition in deployment.  Just don't
break existing DHCP-PD deployments present only where this ND-PD is
tinkered with.

On the IV side -- if the IV expects to be able to provide
connectivity to regular CE routers as well as LVs, then either all
of the CE routers would have to support both requesting mechanisms,
or the IV would have to support both delegating mechanisms
(simultaneously, in case some attaching devices are LV and some are
regular CE routers).

It would be interesting to see who ended up having to do both
mechanisms: vehicles or everyone else. My preference would be that
vehicles should be the ones to incur the additional complexity,
since they're the ones who caused the additional complexity. Of
course, we don't get to choose -- the marketplace chooses. Generally,
the choice is made according to (1) who sees greater benefit in being
able to connect to the other, and (2) who was there first and already
has a significant embedded base. My tea leaf reading is that there's
going to be a huge embedded base of IPv6-connected hotspots and CE
routers with IA_PD long before these vehicles come to fruition. Home
networks and hotspots are already successful without connecting to
vehicles.

I wonder whether any device in home networks has this sub-delegating
capability, I guess not.

For "connected vehicles" to be desirable in the eyes of consumers, I
believe they will have to be able to connect to the consumers' home
networks. So I think vehicles will have to be the ones to go the
extra mile to connect to existing ecosystems.

In a sense I agree.  It seems vehicles are the newcomers when compared
to fixed networks.

However, to make the DHCP-PD work in a vehicle-to-vehicle manner one
would need to modify DHCP (at least make the relationship between Relay
and Server be dynamic).  I wonder whether any work was pursued that way.

My Conclusion: If vehicles are really only ever expected to talk to
other vehicles via this particular link, the proposal could be
considered in the vacuum of its stand-alone ecosystem. But if the
possibility is likely that vehicles will want to interact with other
ecosystems over that same link, then somebody has to suffer and
support both ways of doing the same thing. The savings of a couple
of short messages at attachment time is, IMO, not worth the
additional complexity on either side. And the advantage to vehicles
of being able to operate in both ecosystems is much greater than the
advantage of saving a couple of messages at attachment time. Barbara

I read and thank you for the message.

Alex





--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to