Hacker News Reader

Comment by aarchi | Hacker News Reader

Fortran has a lot of interesting current compiler work leveraging MLIR, a subproject of LLVM: Flang [0] and CIL for C/C++/Fortran [1]. It seems like Fortran tooling is making somewhat of a comeback.

[0]: https://github.com/llvm/llvm-project/tree/main/flang

[1]: https://llvm.org/devmtg/2020-09/slides/CIL_Common_MLIR_Abstr...

View on HN

"I don't know what the language of the year 2000 will look like, but I know it will be called Fortran." —Tony Hoare

I have seen this quote thrown around a lot and have never had the proper context to understand what Hoare was getting at.

The quote is from 1982[1], which seems a bit too late for any “Fortran juggernaut” jokes to have been relevant, and even in 1982 I would have guessed that Fortran was considered a bit old-fashioned. Was he commenting on its comparatively simple design, compared to “flashier” languages like Pascal? Was he just noting its ubiquity in science / government work? Or was Fortran at the time considered a buzzword, something cool-sounding to rope ignorant managers into approving tech projects?

I am a Young Person (thirties) so this era of computing is before my time. More to the point, Hoare’s statement was a deliberately squishy social comment and not a rigorous computer-science comment, which makes it harder for outsiders to “objectively” understand his point. I just can’t see a way that Fortran really fits in the comment - what was the punchline supposed to be?

[1] https://arstechnica.com/science/2014/05/scientific-computing...

In 1982, FORTRAN use was widespread. The limitations of the language were well known, and it was "obvious to all" that the languages of the future would be more sophisticated and overall "better" than FORTRAN. It was not "cool", but very mainstream. Taught to all engineering majors in college. And it was the go to language when COBOL wasn't a match. I interpret this statement as pointing at the tremendous inertia against change, and the fact that most of the tasks given to the computer do not require a abstraction and advanced concepts; just the "IF THEN", "DO", and assignment. I think he was off by 10 years. Even now "You can write FORTRAN in any language". I have seen javascript done this way. And it works. We can argue about how well, but it works.

In 1982, I knew of someone at a university who was working on the Fortran 8x specification that became Fortran 90. I don't remember it being thought of as old-fashioned, maybe just less of a general purpose language than in earlier years.

Since you honored me by referencing my article, I’ll offer my interpretation. Notices he says “the language”. At the time, Fortran was the language for numerical/scientific computing. By 1982 it had evolved significantly from its origins 30 years earlier, but was still called Fortran. He is saying that it will continue to evolve, but no matter what it will look like in another 30 years, it will still be the language, and will still be called Fortran. He was prescient: Fortran 90 included radical advances to the language, but it remained, pretty much, the scientific programming language, and was still called Fortran. The last sentence of the article, which is from 2014, refers back to the epigram, and predicts that Julia may be the language that finally supplants Fortran for use in science and engineering. Now, Fortran is still dominant in high performance scientific computing, C++ has a niche in particle physics, but Julia is, in fact, making rapid headway across a wide variety of disciplines. So I think my prediction is turning out to be pretty good.

There is also LFortran, an interactive Fortran compiler:

https://lfortran.org/

Which has multiple backends, besides the default LLVM one, it also has a C++ backend (to translate Fortran projects to a readable C++), and once MLIR matures, we'll add an MLIR backend also.

(I am the original author.)

Reply on news.ycombinator.com