Richard P. Gabriel: The Art of Lisp & Writing

Lisp vs Java is one thing but there is an undercurrent of "static typing is terrible" running through this paragraph (and others). I am not sure I agree. Haskell for example is very expressive(at least to me). And I am not sure a static type system gets in the way of "prototyping" much.

"Lisp, of course, and especially Common Lisp are my languages. Not since the 1960s have I programmed professionally in a language other than Lisp."

With all due respect to Richard Gabriel, (and this probably blasphemy) perhaps he should? Haskell or Erlang or Factor are mind expanding in ways Common Lisp never was. (Not to deride Common Lisp, but I am tired of people implying it is some kind of unsurpassed acme of language design excellence)

What?

Otherwise, good article.

At most levels of work, we aren't actually making enormous changes all at once - even in the writing analogy, a story has to be written one sentence at a time, and if the concepts start changing partway through, the rewrite has to be done....one sentence at a time. The breakthrough concept that turns everything on its head is usually itself just one sentence.

In programming, we change two lines and then run the program again, we build on known, tested libraries and languages, we develop a concept on paper and then implement it. And because these things are built on a small foundation of logical concepts, we can build the language to ensure ever-greater levels of self-consistency that would otherwise only be manifested at runtime.

Feedback loops.

The fiction writer has a short feedback loop - the "state" of a story is small - no explosions of complexity as one would get with loops or recursion - so its progression can be kept in her head.

The programmer has a longer feedback loop, because her expectation is not to maintain the entire progression of the program in her head, but to let the computer do it and then display the results in some manner.

Dynamicness optimizes the feedback loop for greater incrementalism: the program runs until it hits any one point where an irreconcilable error is encountered, and then multiple options are available to fix the error and even let the program continue.

Staticness optimizes the feedback loop for "batch" error-checking: the programmer writes a full body of code she thinks is correct, with no run/test cycle to speak of, and then gets all the errors at once. When the program compiles, it works.

Since we can't implement either ideology perfectly, it's probably best to have some of _both_.

This is not a very good way of putting it. The essence of engineering is determinate knowledge, i.e. scientific, logical knowledge. That is not to say it excludes other things, or that that essence alone is sufficient, just that it is the definitive, distinctive, fundamental feature.

Experiment produces determinate knowledge -- the most specific, detailed kind. Scientific research produces the broad, general know-how, and experiment fills the gaps for a particular task. If engineering were not fundamentally about determinate knowledge, then experiment could not work. The knowledge produced by the experimental trial would not even have the generality to transfer to the intended use.

Associating engineering and software with art seems of little value. From what I remember of aesthetic philosophy, art does not have substantial, useful, or perhaps any correspondence with engineering -- let alone the vagueness of what art is anyway. One cannot fault the real content of the article, but the arty leaning is ultimately empty.

The limit of engineering is evolution, not art. When there is no knowledge, engineering design can't work, and instead evolution is the way. Proof of the effectiveness of that alternate approach is all around.