Re: [v6ops] draft-binet-v6ops-cellular-host-requirements-00.txt

Hi,

On Nov 15, 2012, at 5:12 PM, Tore Anderson wrote:

> Hi,
> 
> REQ#8 («the cellular host MUST use the interface identifier sent in PDP
> Context Accept message to configure its link local address») is rather
> tricky to implement for a cellular dongle that emulate an Ethernet
> device (such as an USB CDC Ethernet class device). In that case, the
> dongle provides an invented Ethernet MAC address from which the host
> derives an link local address using EUI-64.

And usually you also get other weird stuff along this like the stack
thinking the link really is ethernet and thus requiring address resolution
to work.. which itself is bogus as the 3GPP link has no link-layer
addresses.

> Since this link local address does not have the correct interface
> identifier, ND may (partially) break. In particular, I've experienced
> that when sending an RS from such an EUI-64-derived link local address
> to the network, the GGSN reponds with an RA that is unicasted to the
> link local address the network expects the hosts to have. The host has
> no knowledge of this address, and discards the RA.

Yep. But I say here the UE is more wrong than the gateway since the 3GPP
link definition says the gateway _tells_ the UE the IID to use with the
link-local.

> I don't have a good answer to how this problem should be solved.
> However, if the document could make some recommendations (or at least
> describe the problem), that would be good.

Simple. Get USB class fixed.

- Jouni

> 
> Best regards,
> -- 
> Tore Anderson
> Redpill Linpro AS - http://www.redpill-linpro.com
> _______________________________________________
> v6ops mailing list
> v6ops@ietf.org
> https://www.ietf.org/mailman/listinfo/v6ops