atmosphere - the developer blog of Bernard Wong

Hi, this is the author of the Build Your Own book.

I have often heard this objection. Of course, we may not need Yet Another programming language, but this doesn't address our need for better compilers.

Furthermore, I have worked in the computer business for 30 years, and I have repeatedly heard that we don't "need" various things.

I'll never forget a supervisor at Motorola in 1979 informing me that the developers of the first cellphones didn't "need" a compiler running on their test Z80 (the cellphone was based on the extinct Zilog-80 chip), and that they should be content to use IBM and TSO. She said "all possible compilers are already written".

Of course, this was just as a penniless French immigrant was developing Pascal for micros, etc., etc.

In the 19th century, inventors were told "it can't be done". In the 21st, eager beavers are told "it has been done" and "don't reinvent the wheel".

But even if we don't need anudder programming language (such as the one I'm developing, The Language of the Gods) or another compiler, I have discerned a need for the mere ability to write a parser and interpreter QUICKLY when the user's needs (in law, banking and many other applications) change so quickly, and are so complex, that the user is best served with a responsive way to specify the business rules in real time.

I've solved real business problems simply by using parsing and interpretation as a tactical gesture, and an example appears in my book where I show how to make the owner of Loans for the Honest Poor happy, by letting HIM specify the rules for various forms of microcredit.

Most perniciously, it appears that the organizational Dilbert psychology (gee, it's been done, so we better appoint a study group to stare at the problem until we find a zero cost package, or something) was a factor in Sep 11.

Right, where's my tin hat. But here are the known facts.

Colleen Rowley was an FBI agent on the trail of several highjackers including Zacharias Moussaui, a French citizen dumped from the final team who was detained after Sep 11.

But as Ms. Rowley testified before Congress in 2002, her team was hamstrung in part by the FBI data base they used to track suspects.

They could enter the equivalent query to terroristSuspect, and the equivalent query to attendsFlightSchool. But they could not And (or Or) the clauses together.

When I read her testimony, I wondered why some programmer wasn't appointed to write, in a day, a (1) scanner to parse the basic phrases, a (2) simple, top-down, recursive descent parser, and (3) an equally simple stack interpreter as I'd done in several languages (including Cobol)...without telling the boss.

Of course, since I've worked in organizations like the FBI, I realized that no good deed goes unpunished, and that not everybody is willing like me to try rocking the boat.

I think the basic fear is that if programmers are too self-empowered they will interfere with management. The fear would not exist in a society in which there was less us and them and more partnership, one in which programmers don't have to shuttle between being a partner and being a transcriber of a management "vision" in a crazy fashion.

Compilers are a unifying paradigm across many application boundaries, because many users don't want a "solution" cast in concrete.

They want a language for talking about the solution.

"Business rules" don't have to be code buried in Source Safe (eek) or cvm. They can present to the user transparently as she goes about her business, and the system can even ensure their health in the form of non-contradictory and complete rules through symbolic execution.

Unfortunately for me, however, this idea generates practice and not product, because it moves from code to a world in which "the rules" are known to and assented to by all...not hidden in code. I end up a teacher and not the builder of a better mousetrap who has to pay the rent.

PingBack from http://mydebtconsolidator.info/story.php?id=1834