RE: [16NG] Comments on IPv4CS subnet model and ARP

Thanks for the reply. Please see in-line.

-------->

 [sekim] For example, legacy IP host is a notebook PC with IEEE802.16 wireless interface on Window by Microsoft.

IP hosts send a packet to the Base Station (BS) by unicast. Because IEEE 802.16 does not support broadcast among the nodes.

The uplink, from node to BS, supports unicast.

The downlink, from BS to node, supports multicast and broadcast.

Therefore, [RFC1207] should be used to support ARP capabilities for point to multipoint mode in IEEE 802.16 networks.

[SC>]  In case of IPCS, there is no MAC address involved. A BS should be aware of that. I am not sure why it  would try to do ARPing on the downlink although it is capable of broadcasting in that direction. BS may do ARPing for IP-over-ethernet CS link.  One question remains : what would be the behavior of ASN-GW when a node is on IP-over-EthCS link corresponding to a node with IPCS link ? Should we allow this kind of configuration? This brings in complexity; we can limit a single type of access CS per gateway ?

 If not, legacy host should use broadcast like IEEE 802.3. This means multicast and broadcast scheme are required in IEEE802.16 networks.

[SC>] Not sure what you mean here. Do you mean that a legacy host should be able to use the 802.16 network without any change in the IP/ARP mechanism?  I am thinking this example as today’s implementations to connect as an IP/Ethernet connection vs IP/Point-to-point connection. For point-point case, there is no MAC address used, thus usually the implementation does not do ARP. We can draw the analogy of point-to-point link with IPCS link. 

Before sending the PDU down the IP layer should know where it is sending the data to. If the link is a IPCS link, then it does not require ARPing.  But I understand that the draft is not clear about how/whether a IPCS link should interoperate with a regular 802.3 style MAC device.

Thanks,

-Samita