Modeling Wireless Links for Transport Protocols

This paper characterizes several types of wireless links regarding how they may effect wireless links. The paper seems to be primarily motivated by poor models used for wireless links in prior work. Some areas for serious modelling errors:

• losses — most link-layer protocols use forward error correction with a limit on retries, so loss rates due to errors aren't very high (as often assumed) except in very poor radio conditions. More common, the paper claims, are bursty losses from handoffs.
  
• delay variation — contrary to the default of many simulators, delay jitter is frequently high due to error recovery and handovers and link scheduling (e.g. in cellular links) can cause delay spikes


• not duplex
  
• traffic models — consider multiple flows and flows including more than the wireless link
  

The paper gives suggestions on how to model these and some other effects (e.g. queue management, resource allocation, bandwidth variation, asymmetry) and provides some examples of how varying these parameters of the model can drastically affect the performance of transport protocols. The paper's value really does not seem to be so much in its evaluation of model differences (though they show some non-obvious results: e.g., buffer size can have a substantial effect on transport protocol behavior) but in providing a checklist of mostly simple mistakes to avoid.