This work started out with our discovery of a pattern of throughput distributions among links in IEEE 802.11 networks from experimental results. This pattern gives rise to an easy computation method, which we term back-of-the-envelop (BoE) computation. For many network configurations, very accurate results can be obtained by BoE within minutes, if not seconds, by simple hand computation. This allows us to make shortcuts in performance evaluation, bypassing complicated stochastic analysis. To explain BoE, we construct a theory based on the model of an “ideal CSMA network” (ICN). The BoE computation method emerges from ICN when we take the limit c \to 0, where c is the ratio of the mean backoff countdown time to the mean transmission time in the CSMA protocol. Importantly, we derive a new mathematical result: the link throughputs of ICN are insensitive to the distributions of the backoff countdown time and transmission time (packet duration) given the ratio of their means c. This insensitivity result explains why BoE works so well for practical 802.11 networks, in which the backoff countdown process is one that has memory, and in which the packet size can be arbitrarily distributed. Our results indicate that BoE is a good approximation technique for modest-size networks such as those typically seen in 802.11 deployments. Beyond explaining BoE, the theoretical framework of ICN is also a foundation for fundamental understanding of very-large-scale CSMA networks. In particular, ICN is similar to the Ising model in statistical physics used to explain phenomena arising out of the interactions of a large number of entities. Many new research directions arise out of the ICN model.