#FORTRAN and #COBOL are dead.
Long live Fortran and Cobol.
Modern Fortran is indispensable for high-performance, scientific computing, like weather simulation on supercomputers. Modern Cobol is indispensable for high-throughput, business computing, like financial transaction processing on mainframes.
But Fortran and Cobol suffer from the image problem. Young #programmers will not devote their careers to these seemingly dead languages. As such, many Fortran and Cobol shops are desperately trying to "modernise" their codebases by translating into C++, Java, Python, etc.
This is a mistake. A weather forecast that takes a couple of hours for a Fortran implementation that runs on a 1000-CPU supercomputer will take months for a Python version that runs in an enterprise cloud. Analogous examples abound for Cobol. These niche systems are cloud-proof—they will not bend to the charms of cloud computing.
New language features and implementation techniques are continuously, albeit gradually, being integrated into Fortran and Cobol, and new supercomputers and mainframes are still being designed and manufactured. Yet, there is no injection of new programmers into these specialised domains.
A sensible approach, then, is this. Instead of converting pieces of code written in 60yo languages into those written in 30yo languages, design brand new languages—with dependent type system, algebraic types, type inferencing, memory safety, and other accoutrements of modernity—that target standardised Fortran and Cobol, much like TypeScript and ReScript target standardised JavaScript to "modernise" web development. And if these new languages become established, retarget them to binary.
Add comment