Re: [v6ops] draft-ietf-v6ops-64share

[Vízdal Aleš <[e-mail]>]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of
> Alexandru Petrescu
> Sent: Thursday, January 24, 2013 10:56 AM
> To: [e-mail]
> Subject: Re: [v6ops] draft-ietf-v6ops-64share

Alex,

> > To my understanding there is no link-layer addressing at the 3GPP
> > interface, so it wouldn't be easy to do l2 bridging.
> 
> Well it depends.
> 
> In general I think very many cellular interfaces on the UE do not use
> link-layer addressing.  The devices have 'all-zero' MAC address, use the
> "NO_ARP" flag and so on.

What are you seeing is not the 3GPP link traffic, but the Ethernet emulation 
done by the modem firmware. To be able to sniff on the 3GPP interface level
you would need a tracer to see what's happening there.

> However, more recent devices (e.g. USB keys for LTE: bandrich C502,
> huawei e392, more) using qualcomm chipsets in some instances, using
> 'qmi' interfaces and drivers, exhibit a form of link-layer addressing.
> 
> First, these devices use centrally-allocated IEEE identifiers for
> Company_ID.  Suprisingly to some, the well known file oui.txt exhibits
> such id's for manufacturers of 3GPP-compatible equipment (traditionally
> it mentions only manufacturers of IEEE-compatible equipment).
> 
> Second, some kernel modules, for instance in the 'qmi' family, form the
> last 3 bytes of a 48bit address.
> 
> Finally, wireshark dumps of packets emitted by these cellular interfaces
> show Ethernet headers with MAC src and dst addresses, and interprets
> them according to oui.txt (wireshark interpretation is a good indicator
> about the width of the spread of some protocol).
> 
> These three aspects make me think that maybe the new and future cellular
> interfaces may be more and more link-layer addressable.
> 
> One consequence is the arrival of the use of classical ND on these
> links, and, why not, the bridging forms which are customary on Ethernet
> links.
> 
> (caveat emptor: yes much intelligence exists in the small computer in a
> USB key that may make some things appear what they are not - proxying,
> on behalf of, etc. - yet they don't run the IP stack, it's the USB Host
> computer which runs the stack).
> 
> Alex

Ales