[R-gui] Re: [R] Feedback about SciViews?

Philippe Grosjean phgrosjean at sciviews.org
Tue Apr 29 17:20:47 MEST 2003

Duncan Murdoch <dmurdoch at pair.com> wrote:

> On Tue, 29 Apr 2003 07:55:40 -0400, you wrote:
> >I think that platform independence is extremely important.  You did not
say otherwise, but I was hoping you would stress that point.  -Frank Harrell
> I don't know if I agree with you or not.  I think it's extremely
> important at the level of computations that don't involve the user
> interface.  If you're doing a statistical computation, your functions
> and scripts and packages should work on every platform.
> However, I think it's a hindrance to progress if platform independence
> is enforced in the user interface.


Frank E. Harrell Jr answered:

>Thanks for your note Duncan.  I think that to the extent possible it is goo
>to keep enhancing generic cross-platform tools to accomplish these goals.
>One ulterior motive I have is for people to be able to jump from Micro$oft
>to Linux easily.  A more noble goal is that I encounter great difficulty as
>a teacher when different students use different looking GUIs.



So, this is the problem! Platform independence is desirable, but it is a
major limitation in current GUI development. So what, should we limit to a
tiny set of common features in order to provide same look-and-feel on all
platforms? Netscape, Matlab, Mathematica,... are good example of similar GUI
look-and-feel in different platforms. However, what the story does not tell
is: what was really the effort required to the programmers to archieve this
goal compared to different implementations on different platforms? Matlab
6.x is Java... is this the secret? I have many difficulties to run SJava
under Windows. Once I successfully compile and run it in one machine, once
it fails and I cannot spot the problem. I really wish to get it working

Current developments in SciViews involve more platform independent tools by
1) Web pages. They are platform independent by nature. In the next version,
we will implement electronic reference cards, graph galleries, code
libraries, course and training session asssitant... using web pages and

2) R code. Moving as much code as possible in R ensures execution on all
platforms supported by R. The next version will be indissociable of a
specific SciViews R package, while the current one is managing everything
outside of R. This will have a cost: it will be more difficult to adapt
SciViews to other calculation kernels (Octave, Mathematica,...).

Does the solution go through a mixed GUI with a part of it using platform
independent features, and another part making optimal use of platform
specificities (and thus probably different from one platform to the other)?
This does not solve the question of teaching a course with students using
different looking GUIs... or perhaps different OSes... Kudos to those who
are able to do that (Frank?). I remember the half day I spend to install
S-Plus 6.1 on 25 different computers under Win 98, NT, 2000 or XP. In
comparison, installing R in another course session was sssooo easy, although
some participants were using Windows, other MacOS and one was using Red Hat!
So, I know deceptions that originate from non-platform independent software,
but for GUI, I feel we have no choice currently.


Philippe Grosjean

 ) ) ) ) )
( ( ( ( (       Dr. Philippe Grosjean
 ) ) ) ) )
( ( ( ( (       LOV, UMR 7093
 ) ) ) ) )      Station Zoologique
( ( ( ( (       Observatoire Oceanologique
 ) ) ) ) )      BP 28
( ( ( ( (       06234 Villefranche sur mer cedex
 ) ) ) ) )      France
( ( ( ( (
 ) ) ) ) )      tel: +, fax: +
( ( ( ( (
 ) ) ) ) )      e-mail: phgrosjean at sciviews.org
( ( ( ( (       SciViews project coordinator (http://www.sciviews.org)
 ) ) ) ) )

More information about the R-SIG-GUI mailing list