Forth: The Other White Meat

Forth is metaprogrammable machine language in the way that Lisp is metaprogrammable expressions. Each is a local optimum of programming languages, meaning that any other language that tries to do the same thing ends up being either isomorphic to it or not as powerful. In other words, a version of PG's thesis about Lisp holds equally for Forth in its neighborhood.

But Forth has one major advantage over all other such languages: it's a generalization of the hardware.

Every metaprogrammable language other than Forth is a generalization of some abstract primitive (s-exprs for Lisp, objects for Smalltalk), which leaves a gap at the bottom that must be bridged before anything can be executed. These languages all struggle to get to the machine. Consider the decades it took to make Lisp fast. In Forth, the primitive is the machine. This is so extraordinary a position to occupy -- full metaprogramming all the way up from binary -- that I wonder why we don't see more amazing things done with Forth. Of course, majority practice is not correlated with computing greatness.

I think this extraordinary quality of Forth is connected to how Forth grew out of a working programmer's projects. Lisp came from math, Smalltalk from language research. It's worth remembering that Moore learned Lisp from McCarthy before he invented Forth. Forth is a Lisp hacker's attempt to make something he could use in commercial projects on mid-60s hardware.

I would have said Prolog, it has more history in AI among other things. Forth is like assembly if you have no registers and just a stack. It's still quite low level. If you want mind bendy that you can still reasonably do things with check out Prolog.

The "Thinking Forth" book he likes appears to be available for free:

http://thinking-forth.sourceforge.net/

I used to run FIG Forth on my Apple ][. It was a pretty fun and instructive experience. Not sure if I would choose Forth on larger systems but it's great for resource constrained environments.

Just added forth to my todo-list.. thx!

In terms of comparing Forth and Lisp, I was able to pick up a fairly large chunk of postscript (a forth-ish language) and work with it with no prior exposure to forth.

I haven't seen any Lisp where that is the case. Generally the lisp I've seen gets far too wrapped up in its own cleverness.

My impression is that Forth is a language where things get done, whereas Lisp is a language for basking in your own glory.

I have a theory: the amount of hype for a language is inversely proportional to the degree to which people use it to do real work.

The same could be said of methodology evangelism. The people actually doing the work don't tend to harp on about it.