[R] History of R

[Earl F. Glynn]

"Kathy Gerber" <kathy at virginia.edu> wrote in message 
news:47B733A6.3020203 at virginia.edu...
> Spencer,
>
> I believe this is the first mention of pricing that I've seen.

> Several additional points have been made about the comparison of R to
> Octave, some off list.

> -- Matlab did not alienate developers all that much, so people were not
> driven as much to Octave.

> -- Matlab users often switch to R rather than Octave.

One big reason we use R now was what we considered to be unreasonable MatLab 
licensing terms from the Mathworks. We have mostly biologists here who 
usually have an intermittent need for analysis tools using MatLab --  they 
just don't need a dedicated MatLab license per person for analysis, yet that 
is what the Mathworks expected us to buy. As a new and growing research 
institute, for several years it didn't make financial sense to have a single 
shared network MatLab license that cost the same as four named-user licenses 
when most of our users were biologists.  In our opinion, the Mathworks 
wanted us to pay too much for too little use of their product due to their 
license limitations.
Even though we are a 501(c)3 non-profit research institute, the Mathworks 
refused to give us academic pricing (and still does). About four years ago 
MatLab refused to allow one of our postdocs and me to share a single license 
for casual use. I asked "what is your pricing model"? I asked why the 
Mathworks cared if a post doc used MatLab for two hours a month and I used 
MatLab for two hours a month using the same license. So, frustrated by the 
licensing inflexibility of the Mathworks at that time (four years ago), I 
abandoned MatLab and re-wrote the MatLab project I was working on in R, and 
do most analysis now in R.  I avoid using MatLab as much as possible and 
explain licensing terms to new students and post-docs as we start new 
projects when MatLab is proposed as a solution.

Over the years, the Mathworks has been inconsistent on whether a single 
license on a PC can be shared.  About a year ago, we were told it was OK to 
share such a license on a walk-up workstation, but who wants to walk to 
another floor or building to use MatLab?  The networking option is quite 
expensive, especially to support a number of MatLab toolboxes.  Late last 
year we finally did have enough MatLab use to warrant the purchase of a 
network license, but only for a small number of toolboxes.  MatLab is rarely 
my tool of choice when R is such a good alternative.

Non-academic pricing from the Mathworks for a non-profit research 
environment, combined with the large number of R and Bioconductor packages 
that solved problems of interest (mostly microarray analysis) resulted in 
much more use of R here than MatLab.

Nearly six years ago, SAS also refused to give us academic pricing because 
we were not a degree granting institution.  About a year ago, SAS finally 
granted us academic pricing, but most of the analysis momentum was already 
for the use of R/Bioconductor.

We casually looked at Octave a few times, but there was no strong attraction 
to use it.  Some early tests showed no problems with computations using 
Octave, but showed some annoying issues with graphics that we didn't want to 
deal with.

efg

Earl F. Glynn
Scientific Programmer
Stowers Institute for Medical Research