Same features, a third of the code, and the back button works. Nobody has asked for the loading spinners back.
Restore one file at random, once a month. That is the whole practice, and it is the only reason I know my backups work.
Wp/webdev · 8 days ago
We measured how long our users wait before the first byte, then we measured how long they wait after it, and the second number turned out to be the one that moved the business
Time to first byte was 180ms and everyone was proud of it. Time to a usable page was 4.1 seconds, and every millisecond of that gap was our own code.
Four weights, two of them unused, one of them loaded twice because the fallback stack disagreed with the preload. That was 380KB before a single word of content.
Two developers, one deploy artifact, and a bug count low enough that we stopped tracking it weekly. The part I did not expect was how much of the old code had been about the network.
They are not decoration on a type. They are a claim about which of two things outlives the other, and the compiler is checking the claim rather than inferring your intent.
Nothing was wrong with the containers. I just noticed I was running a supervisor inside a supervisor, and only one of them could restart the machine.
Eight machines, a sound meter and a power meter. The quietest one was not the slowest, and the cheapest one drew twice the power of the next cheapest.
Closed the lid at 30 percent on Friday. Opened it on Monday at 28. That is the whole post.
The HTML page finished before our framework had finished parsing itself. Both of them showed the same eleven rows of text.
Four hundred warnings. Eleven were real, about sixty were style I agreed with, and the rest taught me which lints to turn off in the manifest and never think about again.
It does not. The compositor is holding the key until it has finished restoring the surface, and the terminal never sees it at all.
Rp/rust · 1 month ago
I ported a service from Go and the median request latency barely moved, but the tail did, and here is a week of flamegraphs explaining exactly which part of the old runtime was responsible for it
p50 went from 11ms to 9ms, which is noise. p99 went from 340ms to 24ms, which is the garbage collector, and it is the only number anybody on the support rota ever felt.
The assertion is a string comparison against the exact bytes the browser will receive. There is no renderer to mount and nothing to mock.
A git repo and eleven symlinks. Every tool I tried was a way of not writing those eleven symlinks.
An office thin client with 8GB of RAM. It has served every photo, every note and every DNS query in this house since 2023 and has been rebooted for kernel updates only.