Lambda the Ultimate

• Variable declaration is optional. If you use a variable without declaring it, it's automatically added to the global scope.
• Blocks aren't closures, they aren't even scopes. Instead, functions are a single closure/scope. That results in all sorts of annoying bugs, esp. for people like me coming from C/C++/Java.
• There is no const keyword to complement the var keyword. I would really like to be able to declare constants.
• No access specifiers for members. Everything is public. If you want to make something private you need to use tricks, like making it a variable in a closure.
• No control over which members participate in enumeration ( for ... in ) and which don't. Nor any control over the order of the enumeration.
• If you look at the three previous points together you see that what I really want it to be able to add properties to a value reference. JavaScript only allows adding properties to the value itself.
• In member functions you must use this explicitly to access members, e.g. this.value, I would really like a shorthand notation, say .value
• That fact the ending statements with a semicolon is optional adds ambiguities to parsing.
• I would really like to be able to access closures as objects (like what Sjoerd does with Loell). This would make it possible to externally change values of lexicals, maybe even to play with the closure hierarchy.

Pascal was a teaching language. C is portable assembly. Lisp is very much the lambda calculus. Perl is expressive glue.

Perhaps I'm wrong, but I don't think Lisp is a 'Functional Programming Language' or the 'Lambda Calculus' variety. Yes, Lisp is very much part of the FP Language scene, but that's because it's a language that is flexible enuf to have been part of every programming paradigm undertaken. Oddly enuf, people don't call Lisp an OOP Language, even though CLOS was the first language to provide an standardized OO framework.

The buzz is about Functional Programming Languages these days, and as usual Lisp is able to adapt itself to the paradigm without an inordinate amount of contortion. But Lisp also does imperative, logic, object oriented, etc... The beauty of Lisp is it's ability to be used in practically any programming language paradigm ever thought of. That's why Lisp will still be around 100 years from now.

But Lisp also does imperative, logic, object oriented, etc...

Common Lisp doesn't do logic programming without assistance; Scheme doesn't do OO or logic programming without assistance (of course, "Objects are a poor man's closures." — Norm Adams).

Having said all of that, the observation that Lisp is still the most capable metalanguage, in the literal sense of "language to write languages in," still strikes me as true.

From the online version of the unpublished O'Reilly book on O'Caml: http://caml.inria.fr/oreilly-book/html/book-ora202.html

"Caml, the ancestor [of O'Caml], was developed between 1986 and 1990 by INRIA's FORMEL project in collaboration with University Paris 7 and the École Normale Supérieure. Its implementation was based on Le_Lisp's runtime. It integrated within the language the definition of grammars and pretty-printers, which allowed communication of values between the language described and Caml."

You must be joking!

Toi.

Yes, Haskell and ML people develop experimental languages.

And I was going to add, `No, they do not host them in LISP,' but what I really meant was, `No, they do not write them in LISP.' However, I think very few are hosted in LISP also. I can think of two examples which were, though, and there were once undoubtedly more, but seems no longer to be a popular technique. Yale Haskell (which no longer exists) was hosted in LISP, and an early version of Caml (called Camlot, I think?) was compiled to an early version of Bigloo (which is a Scheme).