@sir If you want a rackmount box, direct from Gigabyte in the US is probably still the best choice. You can call the B2B line and get a single unit. Look for models like the R150-T62.

@sir Not sure what you consider expensive (relative to business expenses) but a ridiculous 96 core, 1TB DDR4 Cavium ThunderX can be around ~$4000 and would likely handle all your needs for a long time. Most annoying thing is actually getting current pricing and most have to be built-to-order, never in stock.

n | █==== | 20%
e | ██=== | 40%
v | ███== | 60%
e | ████= | 80%
r | █████ | 100%

@sir I've found following too many people makes for a very messy feed. Mastodon isn't terribly good at picking "interesting" things to fill your feed with (which is a really hard problem). Ex: Following a magazine results in their posts suppressing everyone else.

@sir @mastohost status.masto.host/ gives a pretty good overview of the setup, I'd be interested to know what hardware it's running on (how many sidekiq processes to an app server?)

:python: Can't believe the fediverse has a haskell emoji but not a python one! Fixed that...

@sir There's a reason I pay some other more schmuck $5/month to do it @mastohost +1

"I just did a simple comparison with jsonpath_rw and pysimdjson is 700x faster haha!"

User feedback is always rewarding :) Makes the weekend spent making bindings 100% worthwhile.

@sir Being able to select cpu features for builds (even if it means taking longer to queue) would be neato and unique. Having issues running simdjson tests on travis since some hosts support AVX2 and some do not (pre-haswell).

@sir What is the retention period for pastes?

We have a new L5 hardware update. Find out what we've been up to for the past few weeks! Big news: dev kit panels are now working with latest update! See: puri.sm/posts/librem-5-hardwar #ChooseFreedom #Librem5

simdjson is pretty neat, performance is on-par with ujson in python-land with trivial bindings. github.com/TkTech/pysimdjson


This is a private invite-only instance. Shoo, shoo.