MACAW: A Media Access Protocol for Wireless LAN's

This is a very well-written paper about media access in wireless networks. Naively, one might expect that the RTS-CTS exchange seen in undergrad is enough to solve most of the problems in wireless media access control, and treat wireless LANs as yet another link layer, but this paper shows how, in many areas, simple algorithms fail to perform, as well as how to fix them. The contributions can be summarized as follows:

• A milder backoff algorithm, combined with sharing backoff values with other nodes, to improve both performance and fairness.
  
• ACK messages for fast retransmissions when loss rates are high.
• DS and RRTS messages to synchronize timing when senders are competing for the medium.
• Separating backoff values by destination, to deal with the problem of noise / local contention causing global performance drops when the backoff is shared.
  

The main themes are sharing contention information to increase fairness and synchronizing senders to prevent one stream from monopolizing the channel.

I really enjoyed reading this paper because it manages to introduce a problem, describe existing approaches, explain problems associated with them, and provide well-motivated solutions in ten concise pages, without requiring background knowledge or skipping over details. It's uncommon to find papers which both define and explore a field this way.

My main criticism is about the choice of setting and evaluation methodology for the paper. The experiments in the paper are all based on simulations, which are in turn modeling the "base station and pad" environment at Xerox PARC. This leads to two questions: First, how similar is the PARC wireless LAN to other types of wireless LANs? Are there issues in other wireless networks that don't arise in PARC's, e.g. due to capture and interference? Second, how well can any simulation capture the variations and subtle effects of wireless transmission in the real world? The paper would have been stronger with an evaluation using real pads, or with other wireless technologies, if any were available at the time. Nonetheless, the problems studied and the solutions proposed are pretty general, so the work should be useful for designing new wireless networks.