[dtn-interest] Re: [manet] Re: PacketBB

[Wesley Eddy <[e-mail]>]

On Wed, Mar 28, 2007 at 03:31:31PM +0530, Ian Chakeres wrote:
> Charlie, I encouraged you (and others) to review PacketBB and to  
> suggest improvements. If you feel PacketBB is too heavyweight please  
> make specific suggestions to the authors and this list.
> 
> I personally disagree about the expensive cost of PacketBB and the  
> cost of using a TLV structure. Can you please provide some specific  
> examples where the cost is large? Or where we might save lots of bits/ 
> bytes for NHDP, DYMO, or OLSRv2? In my analysis PacketBB almost  
> always results in fewer bits/bytes than a non-compacting format.
> 

The IRTF Delay/Disruption-Tolerant Networking RG (DTNRG) shares some of
the same concerns with MANET about saving power by transmitting fewer
bits.  I believe one of the techniques DTNRG is using may help the MANET
packetbb a bit in this case.

One concrete way to shave a few bytes is for the 'L' in each TLV,
instead of using a 16-bit field, to use an SDNV (as done in the DTNRG
protocols):
http://www.ietf.org/internet-drafts/draft-eddy-dtn-sdnv-02.txt
(an update to this draft including C code is in the works)

I haven't looked closely, but I speculate that several of those Ls in
the protocols that use packetbb can fit within 7 bits, and so you should
save a few bytes per message by doing this, at the expense of only
slightly more processing.

-- 
Wesley M. Eddy
Verizon Federal Network Systems