Nobody ever gets credit for fixing problems that never happened (2001) [pdf]

This, but seriously:

Let your junior colleagues fail on a schedule within the margin of error on your planning, but keep a close enough eye on them so you know how to bail it out if they can’t pull it together with a little extra time/guidance.

Okay — still give good code reviews, but if you let them face-plant a little on design, etc, then they get experience. And you don’t look dumb when it turns out they were right. But you do look like a hero when they’re struggling to get across the line, and you whisk in to fix it.

A better way to think about it is that you can trust the code you test. Y2K was a problem because it affected _everything_ because anywhere a date was used had to be checked. For mortgages, yes, banks did fix that code earlier but they did it because someone noticed a problem in 1980 and fixed that specific problem, not because they transformed how they wrote code and checked all of the millions of lines of other code which nobody had reported problems with.

Roll ahead a couple decades, and now you have even more code with lots of interconnected assumptions and the guy who originally wrote it retired to the beach at Margaritaville with zero interest in going back to Cleveland to talk about date math. If you’re lucky, a pile of cash could change that. If you weren’t, they passed in a boating accident last month.

My first tech job was working for a COBOL vendor in the mid 90s and heard a lot from customers and other people in the industry. Much of it was cosmetic (rolling from 1999 to 1900 on a display, etc.) but a lot of it would have had real consequences: not accepting credit cards with dates in the new millennium, calculating interest incorrectly, sending people incorrect notices or failing to send correct ones, etc. I remember a someone at Mastercard saying they wouldn’t have been able to processed credit card transactions at all after midnight if they hadn’t done several years of careful work.

The two things which helped were that it got enough attention to get the business people to pay for remediation, and there were enough dates in the future that people got wake-up calls in different industries. Mortgages were the earliest but closer in you had things like credit card expiration dates which caused enough problems that people got serious about deploying updates.

These days, I’m wondering how 2038 will go. We have a LOT more devices floating around and there are still plenty of new devices shipping with 32-bit embedded Linux which may never get updated. Hopefully most of that cheap IoT stuff won’t last another decade but I’m kind of something like an automotive company getting publicly outed for skimping on technical debt management and some people getting locked out of their smart houses.

It's always worth making some effort at promoting your value; no one else is likely to do it for you.

The same thing occurs for many optimizations. Unless something is unbearably slow and you speed it up 10x or 100x; most people won't notice.

You might spend some time making an often called function return in 1 millisecond instead of 2 or 3 and some might notice that things 'seem' snappier; but no one knows why.

I remember this blog post, it was also a chapter in their book Rework https://signalvnoise.com/archives2/dont_scar_on_the_first_cu...