Love my "COBOL" Google-search emails

“COBOL is back, and AI is writing it. Who’s Verifying the Code?” popped up this morning. And that is the only question one needs to ask.

The blurb points out that COBOL is the oldest [not by a long shot … FORTRAN has it beat] and (arguably) the highest-stakes code in the world. No argument on that second part. It’s not worth belaboring the stakes. Or the challenge.

Yes, the article, should you want to pursue it, is off into the marketing of AI for analysis and drafting and all that. But the arrow in the bullseye splitting all the other arrows in the middle of the bullseye is verification of any result of any “modernization” or “re-architecture”. There is no question that it must be verified to perform the exact function of the original. And, so far, the AI teams are making no such claims and if they did I don’t know who would believe them.

Using the low-end estimate of only 200 billion lines of code in production globally remember that 20-30% was documentation/comments. Personally, I’ll pass on verifying that the 140-160 billion lines of new and changed code does precisely what the COBOL does already.

And remember team, you can just recompile the code, pretty much unmodified, without changes to any of the logic and get exactly the same results we got with the source in the 1970s and 1980s. With GCC optimization and tight integration at the Linux, container, Cloud platform levels. A recompile costs a tiny fraction of a full-on modernization, pays off much more quickly, and has nowhere near the risk of multi-GigaDollar failures.

Besides, the big issue in getting off proprietary COBOLs and platforms is all the other STUFF and the intricacies of databases and data migration. That’s where the effort MUST be made.

Have fun out there! We’re happy to help … still looking for a few brave souls willing to try one of the Open Source Software COBOL compilers we support. Free tech support for limited evaluations/POCs.