[SensorNetArch] draft slides for friday

[Cheng Tien Ee]

I've been thinking about naming a little more. What I'm thinking about 
is more specific to addressing, at the link and network layer.

1. Network layer addressing / naming: The destination address at the 
network layer defines the network destination node. The network layer 
can be divided into 2: the control planes (corresponding to the 
individual network protocols), and a common data plane. When a packet is 
received, the data plane asks the relevant control plane what it should 
do (forward / intercept / accept / multiply). In the case of forwarding, 
the control plane gives the link-level address of the next hop for that 
packet, based on certain fields within the packet (defined by the 
protocol).There is thus a simple interface between the data plane and 
all control planes, and whatever naming scheme a network protocol uses, 
does not affect the rest of the stack. A network-level multicast 
operation can result in the data plane creating multiple copies of the 
same packet destined to different link-level recipients. This doesn't 
make use of the fact that a wireless transmission is always a broadcast 
and can thus potentially be better made use of, but see below.

2. Link-level addressing: we can imagine that the downstream one-hop 
nodes of a multicast tree are given an address corresponding to the 
multicast-group. This notion of reaching a subset of neighbors using 1 
pkt is one of the selling points of Opportunistic Forwarding. The 
downside currently is that we can't individually have each recipient 
ack, so we sacrifice per-hop reliability.

David Culler wrote:

>
>------------------------------------------------------------------------
>
>_______________________________________________
>SensorNetArch mailing list
>SensorNetArch at Millennium.Berkeley.EDU
>http://Mail.Millennium.Berkeley.EDU/mailman/listinfo/sensornetarch
>  
>