We consider primitives for decentralized collision
free scheduling of transmissions. The idea is that each node
updates two local state variables based on the period of activity
that it senses from other nodes. Progressively all nodes are
attracted to partition their periods of activity in the desired
schedule. This methodology is scalable and decentralized, and
is good candidate to be applied to in networks bound to have
very low complexity. We discuss how, with two local clocks
and appropriate update function, the system converges to a fair
schedule, and characterize the convergence speed of the proposed
algorithm