Gerhard Postpischil wrote: <begin extract> The worst problem with English is that words are overloaded (according to one estimate I saw, there are 13 meanings per word on average). </end extract>
This is certainly a problem for English-like programming languages. In other contexts it is helpful. Ambiguity enriches. André Gide devoted many years of intermittent labor to translating Shakespeare's plays into 20th-century metropolitan French. These translations are probably the best available ones, much better for modern readers than Voltaire's; but a red thread runs through his prefaces to them: How can one translate something from English, in which single words have many meanings, into French, in which they are mostly monovalent? Still, he did get the job done well, and the fact that a more pedestrian mind would have found the job harder is not interesting. (There are of course contexts in which ambiguity, semantic enrichment, is undesirable. Notices like «Achtung Minen!» are decisively better than notionally equivalent ambiguities.) Natural languages and programming languages serve very different ends, and it is a mistake to try to make them similar. Mathematical notation in the FORTRAN tradition is a better model for programming languages. One thing can, however, be learned from natural languages. They are context-sensitive. In an example I used here once before the English word 'pen' can mean 1) an animal enclosure, as in 'play pen', 'hog pen', 'holding pen' and the like, 2) a female swan, 3) a writing instrument, 4) a place of confinement, penitentiary/prison, by truncation and with some semantic contamination from 1), etc., etc. In fact, in my experience anyway, context almost always clarifies what is meant. The ambiguity is formal, not substantive. Context-sensitive programming languages that use keywords clarified by context instead of reserved words are very powerful, but academic computer scientists have set themselves against them because they are judged 'too difficult' to implement. They are not. John Gilmore, Ashland, MA 01721 - USA ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
