I wonder how much computing would have advanced if in the 1980s IBM had told their big institutional bank and insurance company customers to get their program code... er... up to code because of the cost of maintaining backwards compatibility. I'd be really surprised if the incumbent legacy industries really were fine with paying IBM through the nose for the privilege of not having to hire people to stay on the upgrade treadmill (and avoid the problem of being stuck on COBOL in 2021...). If not, then that's very, very short-term thinking on their part and you'd think insurance companies of all things would know the true value of long-term thinking...

Well, IBM did kinda cheat: rather than go with Microsoft's approach of making the platform backwards-compatible (e.g. by not fixing bugs that programs depend on, or introducing shims, or redirecting/intercepting calls from known older programs), IBM introduced some pretty hefty OS virtualization technology to their platforms decades before x86 virtualisation was a thing - which I imagine makes back-compat much easier.

Absolutely agree. I'd call it innovation though, not cheating. They had a need and they found a good way to deal with it. Awesome!

P.S.: I still remember when I told my dad I had this cool new thing, VMWare now. There were a few other "yeah, we had that on the mainframe X many decades ago" get-off-my-lawn moments like that :)

Idk, Oracle RDBMS is from 1979, and Java is quite the poster child for legendary backwards compat.

I'll sort of give you Oracle RDBMS. It's also their main product and cash cow.

Java isn't Oracle's though. It was created by Sun (released in 1996) and has had a community driven process for language changes since 1998 (jcp.org). Oracle only bought it and openJDK is a thing.