1.  Avoid notation if possible.  Notation makes it really hard to search electronically.

2.  Put the subject in context, e.g., "In a recent paper, T. Lehrer introduced the concept of left-bifurcled rectangles.  He conjectured no such rectangles existed when the number of bifurcles $n$ is odd."

3.  State your results, in non-technical language, if possible.  "In this paper we show the existence of left-bifurcled rectangles for all prime $n$."

4.  Mention a technique, if there is a new one:  "Our methods involve analytic and algebraic topology of locally euclidean metrizations of infinitely differentiable Riemannian manifolds".

5.  Never, ever, ever, cite papers in the bibliography by giving citation numbers; the abstract is an independent entity that should stand on its own.