One semi safe approach to this is that almost all languages can be written line for line replacement “like COBOL”. There will be COBOL built in functions that need to be reimplemented. There may be internal representation differences that matter.
In the end, there is the same unmaintainable line count code base, and it runs slower, though it may run on faster hardware to offset. This approach introduces the fewest bugs with no real advantage. Rewriting application from modern language perspective is sure to break it much harder.
One semi safe approach to this is that almost all languages can be written line for line replacement “like COBOL”. There will be COBOL built in functions that need to be reimplemented. There may be internal representation differences that matter.
In the end, there is the same unmaintainable line count code base, and it runs slower, though it may run on faster hardware to offset. This approach introduces the fewest bugs with no real advantage. Rewriting application from modern language perspective is sure to break it much harder.
Except COBOL has completely different number handling, so one-on-one migration will cause data corruptions.
Might not mean much on infrequently used systems with hundreds of interactions per year, but on fundamentals like SSNs, this will kill people.
Good programmers usually commit to ethics codes, cowboy AI “developers” usually don’t.