2.9-10    Tightened convergence tolerances. 
	  added survSplit.			
	
2.9-9 more T/F, remove [.terms, depends 1.7.0

2.9-7 Use pnorm5 instead of erf/erfc

2.9-6 two .C calls missing PACKAGE argument

2.9-5 bug fix for identity link in predict.survreg from Therneau.
	
2.9-4 centered= argument was backwards for basehaz (Martyn Plummer);
	add vcov.coxph, vcov.survreg if vcov exists and they don't,
	add anova.coxph[list]

2.9-3 Really change T/F to TRUE/FALSE everywhere

2.9-2 Change 14 uses of _ to <-.  Fixed clogit() to allow missing data

2.9-1 Document the mean->rmean change better

2.9 Change T/F to TRUE/FALSE everywhere, fix segfault in coxph with no
events, make survreg with user-defined distributions actually work and add
examples. Add option "show.rmean" to print.survfit, currently TRUE by
default but may become FALSE in future. Fixed bty option in plot.survfit

2.8 survival passes CMD check for the first time ever (though users won't 
    notice many of the changes). A few little bugs
    here and there. Some better examples added (eg survexp, plot.survfit)

2.7 More CMD check stuff.  HP-UX compilers don't like even static
    variables called 'time' (Peter Kleiweg). 

2.6 documentation/code incompatibilities, three more
sys.frame(sys.parent()) fixes.

2.5 GPL, fixed segfault in frailty
	model with all coefficients NA, predict.coxph standard errors for
	type="terms", new name "survival", predict.coxph
	with newdata but no response <mgri@hal.stat.unipd.it>, bug in
	basehaz, fold in date package

2.4 fixed bug in coxpenal.fit (<m.carvalho@lancaster.ac.uk>) when a
    sparse frailty term is specified before a non-sparse one,
    added clogit(), added Mac makefile (Stefano Iacus), added "Maintainer"
    field, bug in plot.survfit <markc@nswcc.org.au>, better
    environment(formula) handling in coxph, a basehaz() function.

2.3 upstream bug fix for stratifed, penalised coxph

2.2 Fixed anova.survreg,anova.survreglist. Some doc fixes. legends for
    plot.survfit

2.1  options for `n' in print.survfit, [.Surv fixed for R1.1.1.

2.0   New code from Terry Therneau: faster cox regression, bug fixes.

1.1-6 plot.survfit(,fun="cloglog") broke with mark.times=TRUE when the first event was a censoring.
1.1-5 more documentation fixes, added data(stanford2), modified frailty()
   to avoid bug in new save()/load() code.	
1.1-4  documentation and data fixes, particularly survreg.distributions
