Re: 3DLDF

[Hans Aberg]

At 00:33 +0200 2004/08/15, Laurence Finston wrote:
>TeX wasn't intended to be a programming language.

You probably mean that TeX was not intended to be a general programming
language.

>  I use something I call the
>"3 \expandafter rule".  If I need more than 3 `\expandafters' for anything, I
>write a preprocessor instead.  I've found that LISP and TeX work very well
>together.

There are lot of discussions going on how to extend TeX, as it is needed in
the LaTeX3 project and other such. I think that one idea might to augment
TeX with a functional language component, but that is lower down on the
agenda of TeX successors.

>I don't think either TeX or Metafont was intended to be the last word in
>computerized typesetting or font design, respectively.

It is the success of TeX that is its worst enemy: the program was not
designed for the things one attempts to do with it now. Therefore, I think
that one will eventually have to write a wholly new typesetting program,
not upwards compatible with TeX, because that would be a too cumbersome
requirement.

>I admire Knuth's
>accomplishments, but it's his way of going about things that impresses me
>most.  Certainly I  wouldn't want to write a program in the style he used for
>TeX and Metafont, but one could learn a lot by reading these programs.

Try to learn of a lot of the way that Knuth treads, but do not try to
follow in his footpath, because it is his very own footpath not suitable
for others. :-)

>I don't feel anywhere near having plumbed the depths of C++.

C++ is a multiparadigm language. Most users tend to select a certain subset
of features, which they use a lot.

>I'm also an
>applications programmer rather than a systems programmer or a computer
>scientist.  I think the level of systems programming and computer science is
>generally higher than that of most applications programming (although I think
>there's some lousy system programming, too).   I'd like to see the level of
>applications programming improve.

C++ is not a great systems programming language now; but I think it is on
the agenda for the next revision. Some things are very difficult to
implement in C++, most notably a GC different from a reference count
system. This limits its use for implementing other languages.

If one needs an efficient GC, then it might be better to let the Bison
parser output code in a language that has is, rather than C++. (For
example, Haskell compilers has that.)

  Hans Aberg