Message-ID: From: Ivan Godard Newsgroups: comp.arch Subject: some advice to a young man proposing to start a compiler company Date: Thu, 7 May 2020 12:05:35 Lines: 40 Organization: A noiseless patient Spider A decade ago I was asked to advise on a new programming language and it's prospects to be the case for a commercial compiler company. Today he got back to me and enclosed the advice I had given him. I had completely forgotten. Slightly translated, I would give much the same advice to someone with a new architecture proposing to start a chip company. I thought you all might be interested. -------------------------------------- Some general advice: There's no money in language design. All the successful ones got that way being given away for free. There are good jobs as a compiler guy, but no money selling compilers. Syntax is generally unimportant, and should broadly be adjustable to user taste anyway. Semantics matters. Languages must have an easy-to-grasp metaphor at the bottom. It's only worth doing if it fixes a problem that shows up on every page of an average program. Know what else is there. When you can explain to a non-language person why SNOBOL, Prolog and Smalltalk are interesting and C# and Perl are not then it's time to work on your own language. Get back to me when you have a hundred-page self-consistent definition document. Get back to me sooner if it fits in 10 pages. Read the Algol60 Report. Express at least part of your own in Backus-Naur Form. Read the Revised Report on Algol68. Express at least part of your own in van Wijngaarten Grammer. Volunteer to the support committee of an existing language, actively contribute and attend its meetings. You'll learn that languages do not exist in splendid isolation, and language design is a social process with important matters to consider beyond the technical. Be aware that, like most things worth doing, you will spend ten years before you have learned enough to know that what you did was all wrong. :-) Ivan