John and R-ists
While your comments may not be entirely fair to MathSoft, as Peter
suggests, there is one clear benefit of the open source bazaar approach to
writing applications like R or Linux for that matter. This is the
question of verifiability - confused of course by the non-guarantee
comment when you start R (or any other GPL software) and by the fact that,
as you have to pay for proprietory programs, a contract exists that many
feel is a guarantee that the program is correct. In most cases careful
reading of such contracts always reveals get-out clauses that render any
such guarantee worthless.
Consider a bug in a proprietory piece of (closed source) coding. The
checks available are by definition *only* those available to the authoring
company and there is always the incentive to postpone fixes until a major
release is due. With open source, the problem can be fixed by *anyone*
with the expertise rather than anyone with the expertise *and* employed by
the company. The verification *can* therefore be much wider and more
rigorous.
It is rather like the difference between public/academic research and
research carried out by private institutions. Academic research is, by
and large, rigorously checked and verified. Errors do creep in and
sometimes there are open and vitriolic arguments about the correct
interpretation but these are rare. Private research *can* reach such
standards but it is much more difficult for the research teams to convince
their management of the need for rigorous verification if only because for
commercial pressures and reasons of confidentiality. Perhaps the major
exceptions in my experience are where there is a pro-active regulator
breathing down the company's neck willing to close the plant down if
necessary. This concentrates the mind. Unfortunately there is no such
regulator of software - hence the number of fiascos that have become part
of computing folklore recently.
Recent reports from the US are a little encouraging in these aspects.
Major federal agencies - the National Security Council, General Accounting
Office, National Security Agencym Commerce Department - are considering
open source software specifically for security reasons. See
www.usatoday.com/life/cyber/tech/ctg182.htm. Microsoft also appreciate
the power of the bazaar coding method, as seen in the famous Halloween
document (I forget the URL).
John
On Sat, 25 Sep 1999, John Maindonald wrote:
> I wonder whether there is any comment on the following. It is a
> summary of a 10 minute contribution that I made to a biostatistical
> workshop help yesterday in Sydney. My aim was to highlight the
> benefits of the Linux/R development model, in a session with the
> title: "How can statistical packages be improved?"
>
> ==================================================================> The
R Statistical Language
>
> R is a public domain dialect of the S/S-PLUS language.
> Criticism, bug fixes and contributions of code or
> documentation are invited from anyone willing to expose
> their offerings for critical evaluation. Many of the
> well-known statistical computing specialists are actively
> involved in developing R. Contrast the R approach with
> commercial statistical software developers. Typically
> they do not post details of known bugs, they fix only the
> most blatant bugs, and they are remarkably resistant to user
> input. (Stata and Genstat are honourable exceptions.) The
> development model is similar to that followed for the
> development of the Linux operating system.
>
> Most of the independent developers of S-PLUS libraries have
> made their code available for porting to R, or have actually
> ported the code themselves. While R base graphics is good,
> trellis graphics is at present unavailable. Due to the present
> fixed memory model, R may not run very large problems well as
> well as S-PLUS. An interface to a point and click system is at
> an early stage of development.
>
> The ANU Statistical Consulting Unit co-operates with CSIRO
> Mathematical and Information Sciences to offer a series of
> block courses on statistical methodology for researchers.
> The aim is understanding and awareness, not to train
> participants in the use of statistical software. I use R
> for most of the calculations which I demonstrate.
> Praticipants can take my scripts and run them on their office
> or home PC or Linux machine. There has been a strong demand
> to borrow the floppy disk copies of R which I take to the
> courses.
>
> R sources and binaries, and information on history, contributors,
> and ongoing development, are available from the Comprehensive R
> Archive Network mirror sites. The Australian mirror site is:
> http://mirror.aarnet.edu.au/CRAN
>
> ==================================================================> John
Maindonald email : john.maindonald at anu.edu.au
> Statistical Consulting Unit, phone : (6249)3998
> c/o CMA, SMS, fax : (6249)5549
> John Dedman Mathematical Sciences Building
> Australian National University
> Canberra ACT 0200
> Australia
>
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
> 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
>
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._