R-beta: Supporting multiple platforms

[Paul Gilbert]

Frank

Having a fairly large library (about 20,000 lines) of S code myself, I have a
great deal of sympathy with your concern about additional support demands. I
would like to pass along some comments about my experience, and hope this may be
of use and may influence your position.

I have been relatively careful about the design of my library and in large part
have avoided the use of Splus special features. This was a design decision. I
did not like the idea that my investment depended on a proprietary product which
may change and force costly changes on me. (It is almost impossible to stick
with old unsupported versions of Splus. Eventually, operating system upgrades
will force a change.) I have been burned fairly badly by 2 of 4 Splus upgrades
that I have gone through, and am dreading the next one. When I started
developing in Splus one of the reasons I choose it was that I though S was
available and a similar backup was not available for other alternatives (such as
Matlab and Gauss). R has returned us to that more secure position. In the U.S. I
believe the government has a regulation requiring code developed for the
government  be done with languages available from more than one supplier. I
understand it is largely ignored, but the regulation does make a lot of sense.

Porting my code to R helped me recognize a few instances where I relied
unnecessarily on Splus extras. Now that I have completed the port there are only
a few instances where I rely on Splus, for some non critical things which I do
not use very often. I have separate "make" procedures for Splus and R, but one
set of source code and test procedures. So far I have not found that ongoing
support is a problem.

More importantly, certain kinds of error checking are better (or at least
different) in R than in Splus. This has pointed out a few errors which were not
caught in Splus. Also, the fact that I get the same results from Splus and R
gives me considerably more confidence about the soundness of the system.

In situations were I am forced to use "For" loops in Splus I find it much more
expedient to develop and debug in R, then go to fortran or For loops in Splus.

I use Unix, and for me supporting R and Splus has been much, much easier than
trying to help people who want to use my code under Splus for Windows. I say
"trying", because basically I cannot help them. I don't have Splus for Windows,
I don't have any of the necessary compilers, and I'm not really interested in
spending the time. I have relied on others to supply compiled code and explain
the installation procedures. I have found that users are very understanding when
I reply that I can't help them. I think if you are serious about protecting your
own time then you should consider preventing people from using your code under
Windows or Unix, whichever one you don't use. Not that I am suggesting that you
do that, but I think you would save yourself much more time than by prohibiting
its use in R.

Paul Gilbert

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._