Date: May 19, 2013 5:46 AM
Author: Richard Fateman
Subject: Re: Work on Basic Mathematica Stephen!

Rising to the bait..

On 5/17/13 11:38 PM, Andrzej Kozlowski wrote:
> I mention this fact in order to make a point that nobody has been
> making: these sort of things are important for many Mathematical users,
> in fact infinitely more so that the kind basic language issues that
> Richard Fateman has been carping about for over 20 years. These kind of
> functionality is why many people use Mathematica. I therefore absolutely
> disagree with the statement: "I am very happy with the functions
> available in Mathematica 8".

Clearly one can continue to add libraries to Mathematica. Each time
some computational functionality is added 'built in' it replaces a
functionality (or potential functionality) that could be implemented
in the Mathematica language, and may make someone somewhere happy.
Many? I wonder how many people doing statistics are just waiting
for extra features to be added to Mathematica (vs. using systems
that already provide such features, and have done so for many years)..

Possibly some other people will be inconvenienced by cost, size,
complexity, etc. One hopes, not by much, if the design of features
is properly orthogonal to other features. I do not object to
the goal that a computer algebra system should generally be
extensible and extended to cover as much of math as can be
done "nicely". There are some designs that are especially
uncomfortable -- for example imagine the result of merging pre-existing
systems each of which has a different notion of "integer".

My concern has been and continues to be based on the observation
that Mathematica fails to fulfill reasonable expectations on
simple things, for example arithmetic and comparison of floating
point numbers. Therefore
a package built on this foundation of sand may fail in mysterious
(to typical users) ways. A package built by savvy programmers
internal to WRI may fail less frequently than one built for
similar purposes by an "outsider".