[R] Easy, Robust and Stable GUI???

[(Ted Harding)]

On 24-Jan-06 roger bos wrote:
> This highlights the hurdle we face in increasing R users.
> People often use in their career software they learned in
> school and if we can't get the schools to use R (even though
> it free) it makes it all the more difficult to increase R
> usage in business and elsewhere.  If a business person says
> he wants a GUI so he can do some quick analysis w/o learning
> programming, fine, but if teachers and students aren't willing
> to put in a few hours to learn the basics, well, that shows
> its going to be a hard sell.  Just my 2 cents.

Rome was not built in a day, as they say.

I think that the "GUI Question" in the use of sophisticated
(specifically, for us, statistical) software is embarking on
a period of transition.

It reminds me of my early days with Linux (starting nearly
15 years ago). Then, and for a considerable time thereafter
(up to 10 years, depending on what one expected), the dearth
of what people call "applications software" was considered
by many to be the "killer argument" against using Linux.

As someone once asked me "Yes, the idea of Linux is all very
well. But where's the database, where's the word-processor,
where's the spreadsheet?"

For my own work, building on earlier Unix experience, I got
by for a long time using octave (matlab look-alike) for the
computations, and groff (ex troff) for writing documents and
reports. This was the hard way, sleeves rolled up and greasy
fingers, of doing (very effectively) what (already) others
could apparently do much more easily using the applications
(Word & Excel and the like; but of course also TeX for the
sophisticated) even on Windows-3.1 -- but (except for TeX)
with far less capability and quality of finished product,
in my view.

But, especially over the last 5 years, the "applications
software" resources available to Linux have developed
enormously. Now, such old objections are not relevant.

I think the non-GUI approach is currently in a similar
situation that Linux was in 10-15 years ago. We who know
are aware that the command-line allows complete, flexible
and *judicious* exploitation of the full resources of a
package. As has already been pointed out, GUI users are
generally limited to what has been already planted in
the menus, and are also exposed to the risk of *injudicious*
point-and-click: as Philippe Grosjean suggestively put it:
"Just click me!" -- [should there be a tee-shirt to this
effect?]

But I feel confident that, as time passes, the difference
in quality between what gets achieved (in general) with a
GUI, and what gets achieved (in general) with command-lines,
will become perceived. Of course it is possible to achive
good results with a GUI, and bad ones with commands; but that's
my overall view of what the two approaches respectively tend
to encourage.

And (as has also been pointed out), to get a result at all
via the command line requires a certain basic awareness of
the information which must be supplied to a command before
it will work at all. (It is part of the "Gui concept" that
the command which it constructs "behind the scenes" will
be syntactically, and up to a point semantically, acceptable
to the underlying software engine).

While not wishing to discourage development of a GUI interface
to R which will allow judicious and flexible usage, I do feel
inclined to discourage "giving in to the GUI lobby", simply
because that is the only way that such folks know of using
software.

Best wishes to all,
Ted.

--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at nessie.mcc.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 24-Jan-06                                       Time: 14:39:50
------------------------------ XFMail ------------------------------