Considering COBOL was one of the first major languages to be ported to .NET, I would really hesitate to say that is has "no room to grow". Don't get me wrong I have no love for COBOL, but i think that COBOL.NET and all the work done modernizing COBOL during the Y2K "crisis" have actually given it a lot of "room to grow". Sure, the hacker community isn't gonna start writing COBOL versions of Slash and COBOL-nuke or anything, but it a language that was designed for business managers not programmers, and who knows, business managers may cause a revivial (oh lord I hope not), dumber things have been done in the name of commerce.

To say too that FORTH has no room to grow, hmmm, FORTH being FORTH, and therefore being (just about) infintely extensible. I would really hesitate to say that. There are likely more dialects of FORTH out there than grains of sand on a beach. As long as there are electrical engineers and micro-code, there will likely be some form/derivative of FORTH.

Now as for LISP, oh boy, again the language is so extensible that it would never cease to have "room to grow". Take Paul Graham for instance, he wrote the original engine for Yahoo! Stores in LISP (see this for some detail). LISP begat Scheme too, which is used in a lot of places to teach CompSci. LISP (and dialects of LISP) is also the primary language used by Cycorp, which recently got some big DARPA contracts to build "terrorist information databases" and "threat anaylsis tools" (I will not get into politics here, but suffice to say this is some of the coolest and scariest stuff out there right now).

I will agree with you that Smalltalk is kinda "lost in the woods" and seems to maybe have lost some of its old footholds to Java. But it still remains a favorite amoung researchers, so one never knows what will happen.

Languages never really die. They just cease to grow. That's what I meant when discussing Cobol and Forth.

But if they are still alive, and if enough smart people are still using them (sometimes against thier will), they will always grow. COBOL had to grow of we would have been fixing the Y2K problem with punchcards instead of modern IDEs.