Wireshark-dev: Re: [Wireshark-dev] Roofnet Dissector
From: Nicola Arnoldi <[email protected]>
Date: Thu, 14 Dec 2006 13:18:38 +0100
On lun, 2006-12-11 at 13:01 +0100, Sebastien Tandel wrote: 
> Hi Nicola,
>    I've written the first version of the dissector. It only does not
> send data to others dissectors for the moment.
> I've ran it against your capture file and checked some packets.  I've
> seen two roofnet nodes :,, is it right?
>   But ... yes, there is one :) ... see the following
> Obviously, roofnet has several ethernet types. It uses at least 0x0641,
> 0x0643, 0x0644 and 0x0645. It seems like if each of these types
> identifies one roofnet packet type.
> To what I've seen there are :
> - 2 packets 0x0644 identified as data and broadcasted, one for each node.
> - 1 packet 0x0645 identified as a reply
> - a bunch of 0x0643 packets identified as data ... obviously the TCP
> connection
> - and 4 packets 0x0641 with a roofnet type of *0* which is not possible
> with the definition you provide me
> Is it the query type?

The EtherType is modified by the Click router, so don't worry about
that. The Hex value would be perfect! 
> Another thing, looking at the version field. I noticed it was not the
> same for all the packets!
> 0x0643, 0x0644 and 0x0645 = 12
> and again *0x0641* = 4
> Furthermore if length data field seems to be correct ... cksum does not
> seem to be computed for each frame :-/

No, the checksum is not yet computed. Roofnet is still in its infancy
and our implementation is really ... experimental. 

A clarification on the 'next' field.
'Next field' is an integer which tells which of the N hops has to be
considered the next and is updated at each relaying node.

Forward is a link metric in the forward direction on a certain link (you
see that this value is present for each link contained in roofnet
The same happens for rev, which is a forward metric.