matrix07012 @matrix@gameliberty.club

@sjw @nosleep

oh the autozone memes

CHAZ is fyre fest for poor people send tweet

wtf Im not a "channer"
im a gamer

>police simply walked back into CHAZ and retook the precinct with zero resistance
alright that's that then

If you gaze into webpack, webpack also gazes into you

there's something so sad pathetic and poignant about the CHAZ people, they're so alienated from life living in a city that they're trying and failing to do something to help return them to some kind of meaningful Kaczynskian existence by spreading compost over the grass and planting crops in fucking JUNE to give some performative semblance of a connection to the soil. these people are dying inside to become feudal field toiling peasants once again but they don't know how because industry has removed them from their purpose

…… and a ‘Warlord’ 🤦‍♂️

Oh no!
blackityblack.gif

"All software is political, and I think we should acknowledge that"

I think you should acknowledge your need to shut up

All right, here's some prod data on the webvac stuff. (The TL;DR if you missed it earlier: FSE is using venti to serve its static content directly out of its backups. Open-source, here's the README: https://git.freespeechextremist.com/gitweb/?p=webvac;a=blob_plain;f=README;hb=HEAD , here's the README passed through RDoc if you think HTML is a reasonable thing to do: https://www.rubydoc.info/gems/webvac/0.1.5 ).

This data came from the nginx logs for the last 12 hours. They are plots of the time the backend took against the filesize, for GET requests that didn't 404. The filesize is capped at 4MB; anything bigger and we just serve it from the FS.

The first image is a scatterplot with a fitted line for all reqs. It's a little spiky, as you can see, but the overhead looks like O(log(n)), which is better than expected (I figured it'd be linear; probably because traffic patterns mean the same resource is usually requested in succession so the blocks are all cache hits for venti). It is spiky, some of the reqs take up to 10s; this is probably a combination of server load and a cache miss.

The second graph is the same data but it averages time for each individual resource and tosses out the ones that were only requested once. There are still a few outliers, but overall it's performing better than expected.
venti-data.png
venti-data-avg.png