Modeling Wireless Links for Transport Protocol

This paper summarizes a series of issues in current modeling of wireless links and how these inaccuracies can lead to potentially incorrect design decisions for transport protocol. As the underlying link layer characteristics change, the upper layer would need to adapt as well to perform optimally.

This paper characterizes many aspects and details of wireless network that makes it so hard to get accurate model. However, after reading it, I only feel that no matter how careful my model is, it is still not going to represent reality accurately enough. Certainly modeling is much more efficient to evaluate a new protocol or a proposal to change an existing protocol, but all I got from the paper is that modeling is probably never going to be good enough. Now I feel to make a very convincing argument, I will need real deployment data. Perhaps modeling would only serve as a first level filter in a feasibility study.

That leads to my second question. Given limited time to develop a model just to check if my protocol has a chance to perform well, which of these characteristics should I consider. This was not answered in the paper, and I feel that would benefit me and other network researchers a lot.

This paper is still an excellent paper to read, since it summarizes many link layer characteristics in different wireless networks that directly lead to many decisions that we made at the transport layer. It also serves well in the curriculum as a lead-in to a number of more detailed wireless papers.