We discuss a parallel adaptive meshing strategy due to Bank
and Holst. The main features are low communication costs,
a simple load balancing procedure, and the ability to
develop parallel solvers from sequential adaptive
solvers with little additional coding.
In this talk we will discuss some recent developments,
including variants of the basic adaptive paradigm,
improvements in the adaptive refinement algorithm itself,
and a domain decomposition linear equations solver
based on the same principles.