This paper examines packet loss in a 38-node urban 802.11b mesh network, the Roofnet, to understand the causes and implications for MAC and routing protocol design. Key findings include: 1. **Distribution of Loss Rates**: The distribution of inter-node loss rates is relatively uniform, with no clear threshold between "working" and "non-working" links. Most links have stable loss rates over time, but a small minority exhibit bursty losses. 2. **Signal-to-Noise Ratio (SNR) and Distance**: These factors have little predictive value for loss rates. The prevalence of intermediate loss rates is likely due to multi-path fading rather than attenuation or interference. 3. **Multi-Path Fading**: Multi-path fading significantly affects outdoor links and helps explain the intermediate loss rates. Experiments using a hardware channel emulator confirm that multi-path interference can cause loss rates that are hard to predict from SNR alone. 4. **Transmit Bit-Rate**: The robustness of different 802.11b transmit bit-rates varies, with 11 Mbit/s often providing higher throughput than 5.5 Mbit/s even when the loss rate is higher. 5. **Interference from 802.11 Sources**: Interference from other 802.11 activity does not seem to be a significant cause of intermediate loss rates. 6. **Conclusion**: The paper contributes to a better understanding of the causes of intermediate loss rates in wireless networks and their implications for MAC and routing protocol design. The findings highlight the need for protocols that can handle bursty losses and multi-path interference.This paper examines packet loss in a 38-node urban 802.11b mesh network, the Roofnet, to understand the causes and implications for MAC and routing protocol design. Key findings include: 1. **Distribution of Loss Rates**: The distribution of inter-node loss rates is relatively uniform, with no clear threshold between "working" and "non-working" links. Most links have stable loss rates over time, but a small minority exhibit bursty losses. 2. **Signal-to-Noise Ratio (SNR) and Distance**: These factors have little predictive value for loss rates. The prevalence of intermediate loss rates is likely due to multi-path fading rather than attenuation or interference. 3. **Multi-Path Fading**: Multi-path fading significantly affects outdoor links and helps explain the intermediate loss rates. Experiments using a hardware channel emulator confirm that multi-path interference can cause loss rates that are hard to predict from SNR alone. 4. **Transmit Bit-Rate**: The robustness of different 802.11b transmit bit-rates varies, with 11 Mbit/s often providing higher throughput than 5.5 Mbit/s even when the loss rate is higher. 5. **Interference from 802.11 Sources**: Interference from other 802.11 activity does not seem to be a significant cause of intermediate loss rates. 6. **Conclusion**: The paper contributes to a better understanding of the causes of intermediate loss rates in wireless networks and their implications for MAC and routing protocol design. The findings highlight the need for protocols that can handle bursty losses and multi-path interference.