I'd wager it is easier to get rid of C that COBOL. Any takers?
Hmm that's a tough one. Millions of lines of cobol in mainframes running transaction systems, payroll, accounting, etc; huge investment in hardware, financial firms, banks, all the big corporations. Many with lost source code and only binaries available now, but I suppose it originated as cobol so probably still counts as cobol.
Versus... pretty much everything on microprocessors. Windows kernel. Linux and bsd's, the cloud.. Global comms infrastructure - internet routers, satcomms, fibre. TCP/IP stacks. Vast numbers of embedded systems. Automotive. Avionics, some anyway. IOT. Space. Satellites. Military. Medical. Air traffic control. Oh gee.. I forgot mobile. Remembering that many other languages that are themselves written in C, eg Java, so you would have retest everything. There is more: games, AI, EDA, chip manufacturing systems, robotics...
That's a tough call. Hard to say what the aggregate value of either would be, certainly billions of dollars.
Hmm. Training an AI to translate cobol into say... java, sounds harder than training an AI to translate C into say ironoxide. But there's a lot more of the C out there than cobol, and in a much wider variety of myriad deployments on different platforms that would all have to be retested. Just think of all those tesla autopilot highway smashes if you screwed it up.
If financial systems weren't such a central part of this society, say cobol had only ever been used to write chess playing programs, then I would say C wins hands down. But because cobol is still central in so many financial systems, that does even it up somewhat. Certainly it keeps the mainframe hardware refersh cycle going, firms are more prepared to regularly cough up million of dollars to ibm et al in order to avoid having to dump or transfer their cobol apps. However there probably isn't much new cobol being written, and things that were written in cobol are likely to be incrementally moved away to the cloud because it's fundamentally orders of magnitude cheaper. Suppose someone writes a VM emulation of the mainframe hardware that can be deployed at scale onto commodity hardware, and run the cobol code on that emulator.
Maybe in the final analysis it would be cheaper to translate all the cobol into something else rather than rewrite and re-test everything that has been done in C or that depends on software written in C. Or to move that code onto VM's running commodity servers. It hasn't happened yet, people are still buying mainframes. But that doesn't mean it won't happen.
So I think I might come down on the side of C after all. Of course it's cheaper to just keep buying the hardware updates to run the cobol, that just comes out of this year's cashflow, and you know it's a diminishing cost base over time as you migrate off the mainframe.
Sounds like a google interview question....
Final thought: if god said "no more cobol!" and you were forced to write all the cobol apps in other languages, would it be possible? I think the answer is yes. You could do what you do in cobol in other languages. But if god said "no more C!" and you had to rewrite all the C code in other languages... well, maybe, but someone would just have to invent C again and call it a different name with but with a slightly different language design, and call it ironoxide; but it would still essentially be the same thing.
www.theregister.com
old.reddit.com
