Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!mit-eddie!andante!alice!bs From: bs@alice.att.com (Bjarne Stroustrup) Newsgroups: comp.lang.misc Subject: Re: Algol68 Summary: How Algol68 failed to succeed on a LARGE scale Message-ID: <20109@alice.att.com> Date: 24 Mar 91 15:01:20 GMT References: <9168@castle.ed.ac.uk> <46036@ut-emx.uucp> Organization: AT&T Bell Laboratories, Murray Hill NJ Lines: 48 Algol68 was/is one of the most elegant and useful languages I ever worked with. It was the first language where in my personal experience non-trivial programs occationally ran correctly first time they compiled after a major change - this is an effect that I (successfully) tried to achieve for C++. So why didn't Algol68 succeed on a LARGE scale. Naturally, there are many reasons, but here is a few I think were important: - Algol68 was a decade in the making before implementations were beginning to appear - and for many (most?) machines implementations has yet to become widely available. By the time they appeared the initial excitement about the Algol68 concepts had died and early fans become disillusioned and moved on to other languages. - Algol68 developed a terminology and a style of language definition that was clean and surperior to any alternatives I have seen. This made it unapproachable to the casual user. The very elegance of the terminology and style of presentation was a major barrier. I was lucky enough to get my hands on a concise (about 150 page) and very readable description - and a good implementation - otherwise I would have had to dismiss Algol68 as a practical tool and have left it as a ``probably interesting intellectual challenge'' that I would never have had time to look into. - By the time Algol68 was complete and beginning to become available the issues of encapsulation and modularity had come to dominate many people's concerns. Algol68 did not emphasize that - and the textbooks I saw and (with notable exceptions) the Algol68 programmers I talked to completely ignored that and concentrated on explaining how Algol68 was a better Algol or a better Fortran. Thus, by the time one could use Algol68, it was seen as an oldfashioned language - the final close-to-perfect development of a dead branch of evolution. - Algol68 was seen as European and Academic. Either could in itself - at the time - kill a language in the parts of US industry that set the tone for what tools programmers actually use. - Algol68 had a main-frame bias at a time where first mini-computers and then micros emerged as the major growth areas for programmers. I think Algol68 was and is a most attractive language. Had there been a K&R for Algol68 and a cheap and easily portable implementation of Algol68 available for a PDP11 in 1975 or so it would have swept the world covering the areas where Pascal and C thrived. However, there wasn't such a book and there wasn't such an implementation - and it is doubtful if there could have been. The culture that produced Algol68 (in my impression) simply didn't think that way.