Version 2.0-6 2003-09-12  Avoid generating lty=NA.

Version 2.0-4 2001/10/10  tree.q did not work in R, minor improvements
                          to documentation.

Version 2.0-3 2001/08/08  QA changes, e.g. T -> TRUE in scripts.

Version 2.0-2 2001/06/12  Many help-file improvements.

Version 2.0-1 2000/12/08  Version 2 of sm, sm.rm(optimize=TRUE) works.

Version 1.0-3 1999/08/15  Moved .sm.home to package:sm, made ts examples work.

Version 1.0-2 1999/04/02  Modified for 0.64, using R's chull, jitter,
			inst/* mechanism, data/00Index, file.* functions.

Version 1.0-1 1999/02/20:

I altered plot.density (which causes a name clash, and is not a plot method)
to smplot.density.

All the example scripts work except those depending on gam:
	trwlgam1 trwlgam2 mackgam trwlgam3 smackgam    

(and the gam in mgcv does not work with these).


R discrepancies:
===============

is.matrix is false for data frames: I used isMatrix in zzz.R instead.

Colours are (mainly) named in the R version.

In polygon() border gives the colour, not a logical for plotting.

points() does not use the current cex and pch in R.

Labels formed by deparse(substitute(x)) need to evaluated before x is altered.

persp() has fewer arguments, and the default viewpoint did not seem a
good choice for these plots, so I altered it.

glm.fit() has different arguments, and the X matrix must have column names.
Further, weight-0 points do not get the linear predictor set.

is.na(x) fails in R if x is NULL.

Some functions used `break' where `return()' was meant: this does not
work in R.

The uses of unix() and assign() needed revising.

detach() named objects seemed to fail, so I used pos= instead.

BDR 99/02/20
