Show newer

@ShadowJonathan@tech.lgbt why can't we have nice things....

@ShadowJonathan@tech.lgbt they're perfect for the server-timing response, because i'm streaming content so I have no idea how long parts took until they're done sending

@ShadowJonathan@tech.lgbt since http/1.1 but browsers mostly ignore them.. .

and nginx fuckin strips them

fastly.com/blog/supercharging-

ffs, Firefox only accepts the server-timings trailer over https

and nginx does not support trailers on proxy_pass requests aaaa

Show thread

ooh the Server-Timings header is a great use-case for trailers (as opposed to headers), going to experiment with this a bit

this is running in development mode, but on my homeserver which is considerably more beefy than the vps it will run on (because faster network will be more beneficial overall)

Show thread

used a postgres query to find the most replied to toots (and thus hopefully largest threads) on this Mastodon instance

Happy to see my re-ordering algorithm for proper threading, and my rendering still handles them really well. Only the very largest thread took a bit more than a millisecond, everything else is <0.3 ms

Mastodon takes multiple seconds on the /context responses though yikes. Really happy with my stale double-rendering approach so you still get to see content almost instantly, while it's waiting for verification from the Mastodon server that the data is still up-to-date

meds oops 

turns out methylphenidate doesn't do much if you just leave it on your desk instead of taking it

has anyone ever done mood analysis on lastfm/librefm feeds? i dont have one but it'd be so easy to see when I'm not feeling great based on music choice lol

@tobi nothing a whole bunch of stimulants can't fix (i hope)

@jonny i guess? but we already have *two* config flags that are supposed to disallow public, unauthorized api access, so why does gargamel keep punching holes in them

Show older
Pixietown

Small server part of the pixie.town infrastructure. Registration is closed.