[SensorNetArch] Plan

[Cheng Tien Ee]

>If you look at the current text, we've deferred address filtering to
>network protocols to permit snooping, etc.
>
I'm looking at 2.1, which says that reception is dispatched to 
associated network protocol, which is identified by its AM id. I agree 
that packets with the same ID can be filtered by the corresponding 
applications, but, for instance, would it be possible for a snooping 
protocol, with id 8, be able to snoop on packets meant for id 9? Am I on 
the same channel =) ?

>  That way a network
>protocol can maintain any number of addresses.  There will need to be
>some coordination with the link on how those addresses defined or
>transfered up the stack, which we haven't come to a consensus on (note
>the culler remark in the text about naming)
>  
>
Would that be policy we're not sure about? That would affect what the 
addresses actually mean, but the mechanism can be implemented with the 
table I mentioned?

>On Apr 6, 2005 9:43 AM, Cheng Tien Ee <ct-ee at eecs.berkeley.edu> wrote:
>  
>
>>>We have said that every node responds to its address and the broadcast
>>>address.  this is certainly a common link characteristic.  What does
>>>this mean for a multiple name per node model?
>>>      
>>>
>>Don't know if this is way off, but does it make sense to have a separate
>>table, with entries that can be added / removed by network protocols.
>>The columns of the table might be
>>
>>1. Link
>>2. Address
>>3. Network protocol id
>>
>>When a packet is received via a link, this table is consulted. For all
>>matches in the table, then the packet is passed to the network protocol
>>with the particular id. The address field is simply 16-bit (currently)
>>numbers, it's interpretation is up to the network protocol. For data
>>centric storage, the address field would carry the address of the
>>immediate next hop, and at the network level, the final destination
>>address is handled by dcs routing, not SP, so that won't come into the
>>picture here. For a debugging / snooping protocol, it can set the
>>address and link fields to be some equivalent of '*', and SP passes it
>>all packets that come through.
>>
>>e
>>_______________________________________________
>>SensorNetArch mailing list
>>SensorNetArch at Millennium.Berkeley.EDU
>>http://Mail.Millennium.Berkeley.EDU/mailman/listinfo/sensornetarch
>>
>>    
>>