Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!seismo!dimacs.rutgers.edu!aramis.rutgers.edu!paul.rutgers.edu!gaynor From: gaynor@paul.rutgers.edu (Silver) Newsgroups: comp.lang.misc Subject: Re: The Universal Language (Was Re: Efficient Fortran) Message-ID: Date: 12 Aug 90 22:01:48 GMT References: <23950@megaron.cs.arizona.edu> <26920@nigel.ee.udel.edu> Organization: Rutgers Univ., New Brunswick, N.J. Lines: 46 carroll@udel.edu () writes: > I am personally of the opinion that a "Universal Language" is an impossible > goal. Dang pessimists... > ..., a language is a description of a machine, and the form of the language > describes the form of the machine. When you choose a language, you're choice > is based on which language define the machine that most clearly and easily > models the solution that you have in mind. Machine, work upon thyself! > Some models are irreconcilable. For example, try to unify the Smalltalk > model, where object identity is central to the language, with ML, where > immutable data is central. You can't unify them without destroying the > central concept of one of them. Ok, they can't be unified, the programmer must choose between the two. The former model is more common (that of references to data), so would probably be default. Then write a package in the language that modifies it to the latter. When I wrote that such a language must be simple and very versatile, I meant to the extent that it is able to reach into itself and modify its `guts'. Machine, work upon thyself! >>> [All the heat concerning someone@somewhere...] >> And while we're at it, you always put in two signatures! :^D > And frankly Andy, you could could use a shave. 8^) So I did. Fixed. > If you read your proposal, you'll see the error: "The components of the > language should be objects in the language {\em and easy to modify}." What > about language models where objects cannot be modified? Addressed above. > One language can never possibly fit all models together. Not simultaneously, as you've shown! > But a system which allows multiple languages to interact is possible. Sure, this is the current situation, to some extent. What I envision as a UL would be a small kernel language where all these high-level features are provided as packages in the language which embellish or modify it. >> The rest should be provided by library routines. > Or perhaps, library LANGUAGES. Library features, where a feature is a linguistic feature. (My reference to `routines' is unfortunate, it is not what I meant.) Regards, [Ag] gaynor@paul.rutgers.edu