So everything he misses is there. It's just that running legacy code is too risky these days unless you are completely isolated from the internet. And modern infrastructure has a lot more essential complexity.
>It is amazing how well things work – the modern assembler, linker, and debugger handles the code generated by GCC 1.27 without any problems.
https://kristerw.blogspot.com/2019/01/building-gcc-127.html
And the write-up he based his experiments upon:
https://miyuki.github.io/2017/10/04/gcc-archaeology-1.html
I'd say this kind of source-based backwards compatibility is superior to what's happening in Windows world -- not only it encourages you to keep the source code open, but you also don't have to keep as many hacks in the kernel/system libraries.
Then you get laid off at 50 and have no hope of finding another job at the same salary.
Learning `vi` in 1979 then using it for 40 years seems like a good return on investment.
Both would also be examples of survivorship bias.
It isn't as flexible as a PC, but you shouldn't expect it to. You learn the design, conventions and idiosyncrasy of the platform and work with them, not against them.
Maybe this isn't the general standard, or maybe this wouldn't work for multinationals or companies/workforces with different values, but it works here. I suppose it also depends on what your company is doing; we're in fast moving online retail, if you are in a slower operating environment (i.e. only doing batch processing or bulk messaging with ETL of different kinds) where changes aren't giving you an edge over competition if must be a different set of operational requirements too.
Certainly not the i. There is so much documentation that it might be more of an issue of it being overwhelming. there are multiple active communities sharing code.
this is a system which is based on DB2 meaning SQL in native, development tools are varied, from COBOL, RPGLE, JAVA, and more. You can web face nearly any existing process, node.js work is one area we have lots of current work. no down time for reorganizing files, indexes always maintained, and tools open to all on the platform to optimize sql statements. I have developers using all this and more. most use an eclipse based platform (RDi for Cobol/RPGLE) or eclipse itself to work.
admin wise it is a breeze, no need to hunt down a tcp/ip stack, db tools, schedulers, or such. its all built in.
Legacy systems are only limited by the imagination of those using them. the reason the systems persist and grow is because they are not frozen in time and their architecture readily adapts to new technology because they were designed to not be locked down with an inflexible design.
Case a) Lots of years ago IBM changed CPU architecture on their System i. Every program we made was transparently recompiled to the new instruction set on first run, and just worked after that.
Case b) Legacy technologies are taken very seriously. Decades old programs made in obsolete versions of COBOL can still be compiled without issues on the modern systems. No unexpected side effects, no new bugs, it just works. It provides a comfortable path to gradual upgrade.
Case c) The tech stack builds upon itself, which means you can devote years to learn the stack and still enjoy the fruits of your effort for decades. You just have to keep up with the changes, not relearn everything every five years.
Case d) The manuals and reference are exhaustive and absolutely superb, you can become an expert just by reading them. How many questions about mainframe development are created daily in Stack Overflow?
Their biggest drawback is their cost. Mainframe as a Service exist, but it needs to be affordable enough to run cheap experiments, otherwise it won't be attractive to startups.