This is quick list of some possible fixes for the next revision of quantreg

1.  Adapt Khmaladze stuff to handle "br" output.

2.  refine the choice of eps parameter in rq.fit.fn  (this has been a
perennial problem.  As several examples show, it is sometimes possible
to get into trouble with the current 1e-6 tolerance with rq.fit.fn
returning NaN's while a  more relaxed tolerance say 1e-4 produces a
good answer, verifiable from "br".

3.  The trim object in rq.test.khmal could be improved.

4.  Now location-scale and location tests need to be done independently
and should be organized to do both together.

5.  Write anova function for rq objects. [Done in v 3.09]

6.  Resolve some packaging issues regarding linkages to SparseM
and to incorporate rqss stuff to akima, tripack, etc.

7.  Allow bootstrappping routines to do subsampling. [Done August, 03]

8.  Allow percentile methods for the bootstrap methods.

9.  Add predict methods

10. Revise the Khmaladze code which was withdrawn for renovation on March 12, 2004.