I’m not sure if I’m allowed to promote my own (relevant) articles on here, please let me know. I have a series of blogs about my takes on technology and a work in progress series on how to get computers to count really fast.
You must log in or # to comment.
The way I try to make apps snappy is to choose time over space 90% of the time. Yes, I have had php processes that hit 1GB but I’d rather it gets done fast
You’re making a conscious trade off, which is better than most of the heavy apps out there, because they are slow and heavy. It would be interesting to document your decision making in some kind of blog post, especially if there is real hard data behind the decisions. That’s where the engineering lies.
I’ll be honest as a 1 man operation, my flowchart is fairly simple:
-Can psql do this? Then make the statement to do it
-Gotta do it on code? Make sure you have distinct read – process – save chunks so you can see whats slowing you down and minimize cache misses.
I could easily get by with 400MB of RAM, 3GB of storage, and a Pentium III for web hosting, as long as I’m not getting more than… a *couple hundred requests per second
Actually, a electrical toothbrush is enough for that.
No way electrical toothbrushes have hundreds of megs of RAM. I thought that they were tiny embedded systems with an ATMega-tier CPU at most, if at all more than a switch, battery, and motor.
There was also one on Hackaday recently, that runs a website from a solar powered ESP32.
And yeah, hundreds of MB RAM is also news to me. Quite a bit overengineered.
Personally, I prefer it when authors promote their own work – it’s nice to be able to ask questions about it directly.
I liked your article, personally I like seeing authors able to see discussion of their work too, so in my opinion self promotion here is good.
I do think we are trying things with software now that were impossible before, which also plays a role. But then again, RollerCoaster Tycoon 2 is basically a perfect game, and it’s written in incredibly performant assembly code, so we should still ask for more if RCT2 can show us near perfection years ago.
RCT and RTC2 are pinnacles of optimization but I understand why we don’t code in assembly anymore. Still, C++ with a little care can still be very fast.
Scroll through your typical node_modules directory without learning a little something about software bloat. Yikes.
It’s quite a lot, what we expect from our technology now. But we made it this way because the marketplace has deemed there must always be a winner and a loser, so it’s a never ending game of accelerationist oneupmanship.
The market pressures the competitors, the competitors pressure the engineers, the engineers pressure each other to deliver faster and faster. Sometimes they’re backed into a corner and have to focus on more speed and efficiency, which is shortly thereafter consumed by frameworks, languages, and operating systems that are also competing for adopters, and thus supply stuff like JIT compilers and UI frameworks.
Even before we were plunged into the hellscape of vibe coding, you could knock an app together with a kit of parts using a pinch of glue code, having no clue what’s happening underneath the gui. Who cares? My Mac at idle is running hundreds of processes, it can take it. Until of course it can’t.
Back in olden times, a piece of software was painstakingly hand-built in assembler and C over a course of many months. But ain’t nobody got time for that when your manager can shit out an app with Claude in an afternoon.
