Tuesday, November 18, 2008


Traditionally network protocols have been designed with assumptions about delay and loss not being excessive and an end-to-end path more often than not exists. DTNs violate these assumptions and the authors give a list of example scenarios in the genre of mobile ad hoc networks. My doubts about DTNs "becoming important" persist - the scenarios listed here seem outlandish and I am not convinced these are becoming prevalant or will ever. This reminds me of the din surrounding the work on MANETs...all exciting research problems but (I apologize) contrived scenarios. The standard approach of using a special proxy/gateway to connect to these networks along with specialized protocols internally seems to be more than enough. Their argument about not being able to use these networks as "transit" doesn't seem to convince me. Don't get me wrong - I totally understand that the problem is immensely challenging in the context where these networks operate. Anyways, let's read on...

The paper does a good job of listing the challenges DTNs pose and how they are markedly different from traditional networks. This paper was published in 2003 and I am not sure if there was enough of deployments and data, but I would have loved to see some data to back their challenges - what are the communication patterns, how often, how many bytes etc.

Their architecture is pretty solid and their naming and forwarding schemes are nice. The best thing seems to be that this is entirely generic but then it really isn't optimized to any network. And my guess is that the moment you start optimizing it for different scenarios, we will end up at a state not too different from where we are now!

This is a good overview paper laying down a nice framework with all the possible issues (though sometimes it goes overboard, e.g., with security - now that clearly is totally unnecessary in these settings!). But I don't think this paper finds a place in a class teaching networking fundamentals.

No comments: