4 ms·

Not a software dev, but I would say that some code written in something like Ada and working off a floppy has about a bazillion less leaky abstractions than the

Not a software dev, but I would say that some code written in something like Ada and working off a floppy has about a bazillion less leaky abstractions than the new systems which run an entire Linux system (the brand new US littoral combat ships did this) with who knows all what. The former runs for decades without any issue. The latter basically makes maintenance a never ending cost and chore. It is similar to when my father's company switched from a mainframe that ran all their stuff and was written in the mid 80s to brand new off the shelf software. The replacement software is all about trying to make your needs fit into how the vendor's product works. His company unanimously believes the original software was significantly superior in terms of issues (it basically just ran for decades with almost no issues). The issue is the older tech keeps getting more niche, so you eventually kind of have to migrate sometimes.

fy20 2y ago
I guess the reason why this is the case is that code at that time was much lower level, and often had exclusive control of the CPU. I've toyed around with microcontrollers before with Assembly and C, and I found it fairly easy to hold the entire mental model of what the system (registers & memory) is doing at each instruction. I feel this can lead to less bugs, as you can fully understand what each step is doing. Compare that to something like JavaScript running in Electron, where at basically every line there is a chance that something could return an error or take away control of the CPU messing up your time dependant code. Of course there are ways to handle that, but it's a lot more complicated compared to a CPU your code runs on exclusively. Or you just run into bugs with third party code that you need to work around, or update a library and find a patch version introduced a breaking change. This is probably why PLCs are so popular in industry. You can make sure a tiny part of the system works correctly and is bug free. Then as long as the contract of how it communicates & interacts with other parts of the system is defined properly, the system as a whole will operate correctly.

fy20 2y ago

I guess the reason why this is the case is that code at that time was much lower level, and often had exclusive control of the CPU. I've toyed around with microcontrollers before with Assembly and C, and I found it fairly easy to hold the entire mental model of what the system (registers & memory) is doing at each instruction. I feel this can lead to less bugs, as you can fully understand what each step is doing. Compare that to something like JavaScript running in Electron, where at basically every line there is a chance that something could return an error or take away control of the CPU messing up your time dependant code. Of course there are ways to handle that, but it's a lot more complicated compared to a CPU your code runs on exclusively. Or you just run into bugs with third party code that you need to work around, or update a library and find a patch version introduced a breaking change. This is probably why PLCs are so popular in industry. You can make sure a tiny part of the system works correctly and is bug free. Then as long as the contract of how it communicates & interacts with other parts of the system is defined properly, the system as a whole will operate correctly.