[johnnyworker]

Usability can only ever be as good as performance allows it. I don't care where the buttons are when it takes 5 seconds before they show up, or how pretty font is if the scrolling is stuttering, and so on.

From the user perspective, it really makes no difference if the same UI was hand coded in assembly or in something much higher level that is still fast because it doesn't work against the CPU, RAM, storage and other things.

[eikenberry]

There are at least 2 end users for any code, one who uses the product and one who has to work with the code in the future. The second one will care.

[morelisp]

Fuck the second guy, their entire job is to serve the first. And having been that second guy a large fraction of my career, fuck the original author’s OO GoF hexagonal Clean Code too.

[loup-vaillant]

This comment has a point: there are much fewer developers than there are users, and our moral weight is proportionally tiny.

[fulafel]

I think the long tail of user bases makes a big difference here. Most programs have few users. Most devs work on programs with few users.

[morelisp]

Most devs work on programs with orders of magnitude more users than devs, even for “few” users. The absolute number does not matter.

[johnnyworker]

The crap performance comes from bloat, not from too simple code. I also see precious little maintenance going on, at least compared to jumping ship to something even shinier and with even more eternal tech every few years. Microsoft promised Win10 to be the last one and then instantly changed their tune because of pure business reasons.

And some code gets run on so many devices, by so many users. If starting a phone call took 0.1 seconds longer for all humans on the planet for the next 20 years, even a huge team saving 50 years of their life by cutting a corner and not reducing the delay each would have destroyed more than they gained. And everyone of us uses thousands of programs, millions all in all, and if they all started to prioritize the resources of their makers over yours, it would outpace whatever you can as an individual gain by outsourcing your costs in the same way.

Of course I'm not saying spend twice the time to squeeze out 1% more performance for an installer that's run once. But we are not at risk of that anytime soon; instead we're more in a situation where some people people ship or advocate things that are 10-100 as big as other programs doing the exact same thing, and are ass-backwards and regressive in so many ways. All these github repos with 20 config files and dozens of folders many levels deep, with the "actual" code fitting on a napkin if you added it up, plus a "one-line" install that kicks of a cascade with a gazillion possible failure points. [1]

Partly because so many people flat out don't care, as if actually knowing what is going on under the hood, how the words in the text editor translate into a process, and the wider world it's embedded in, was somehow beneath them (and then they turn around and correctly complain about their managers who treat them like they treat their machines and users). But I think mostly because of this desire to re-invent something badly with a lot of fanfare and in doing that help keep the better thing that exists obscure (especially if that thing is free). That's the majority of professional software development it seems, so I don't worry about which of the laid off people will maintain the discontinued products too much.

[1] Remember NASA World Wind? They had a perfectly fine desktop app before Google Earth existed. Now they have a nodejs thing, so I thought okay, that is something I'll install node for. Alas: https://i.imgur.com/G4mQxjb.png <-- that's the "installer" segfaulting and leaving everything as is. I haven't seen something that bad in like... ever.

And I don't care that I can just roll back, or should have done it in a VM or docker. This, to me, is a joke, and I am not interested in fixing it or using a VM image or anything. I wish the people who maintain that sort of stuff the very best, I think they'll need it.