The design of efficient IEEE 802.11 physical (PHY) rate adaptation algorithms is a challenging research topic and usually the issues surrounding their implementations on real 802.11 devices are not disclosed. In this paper, we identify and evaluate the key parameters to design such algorithms. We then present a survey on existing PHY rate adaptation mechanisms and discuss their advantages and drawbacks. We also propose three new 802.11 PHY rate adaptation mechanisms, named adaptive auto rate fallback (AARF), closed loop adaptive rate allocation (CLARA), and adaptive multi-rate retry (AMRR). AARF, proposed for low-latency systems, has low complexity and obtains similar performance than RBAR in stationary and non-fading wireless channels. CLARA is a culmination of the best attributes of the transmitter-based ARF and RBAR control mechanisms with additional practical features such as adaptive fragmentation to improve multipath-fading channel sensing and to provide feedback control signaling. AMRR is designed for high-latency systems; it has been implemented and evaluated on an AR5212-based device. Experimentation results show more than 20% performance improvement in throughput over the default algorithm implemented in the AR5212 MADWIFI driver.