Re: [v6ops] draft-binet-v6ops-cellular-host-reqs-rfc3316update

[<[e-mail]>]

Re-,

Please see inline.

Cheers,
Med

________________________________
De : Lorenzo Colitti [mailto:[e-mail]]
Envoyé : vendredi 21 septembre 2012 09:52
À : BOUCADAIR Mohamed OLNC/NAD/TIP
Cc : IPv6 Ops WG
Objet : Re: [v6ops] draft-binet-v6ops-cellular-host-reqs-rfc3316update

On Fri, Sep 21, 2012 at 4:37 PM, <[e-mail]<mailto:[e-mail]>> wrote:
1. Did you consider a requirement to support RFC 4191? Many people are asking for the ability to support more-specific routes, especially in the MIF working group.
[Med] We didn't considered it because there are some assumptions to be made: e.g., do we expect all interfaces are connected to networks managed by the same administrative entity? How to manage conflicts if distinct policies are sent? etc.

I don't understand. An interface is connected to only network, and the cellular link is point-to-point. By definition, there can only be one entity on the other side. Where's the conflict?
[Med] Consider the case the MIFed device is connected to both a 3G network and WiFI hotspot simultaneously: if the hotspot and the 3G are managed by the same provider, then policies to offload (some of traffic) can be enforced using specific routes. But if these networks are not managed by same entities and each of them are sending specific routes, wouldn't be there potential conflict in the policies? What would be the behaviour of the device?

 2. REQ#28 says the device MUST (no less!) support ND proxy. I don't think it's appropriate to say that an experimental RFC is a requirement. Additionally, ND proxy is not fully baked, and it has issues with certain topologies. We need a better solution than that.
[Med] RFC4389 is the best reference we can quote at this stage. Do you have a pointer to an I-D where these issues are discussed? We can add a pointer to that I-D.
There is no working solution at this time. This can be made to work when tethering over the cellular link, but the ND proxy RFC is not the way to do it. I think we need a new document to specify this.