To the program committee

This paper is a revised version of a paper that was submitted
and rejected by SIGMOD.
The three main reasons for the rejection were:

1. No comparison with DeWitt and Chou's DBMin algorithm.
That is included here.

2. The statement that his algorithm does not do much better in terms
of page replacement tha LRU/2.
We state repeatedly that our main intention is to present an
algorithm with less overhead and about the same page replacement
properties.

3. Finally, we have included further experiments.
