Mima-sama
#Pleromatan stealing gigabytes of your storage for its neverending database hunger :pleroma-tan_corndog:
#Pleroma
RE: https://mai.waifuism.life/notes/9pag98xnswua00a4
@mima never had an issue with this actually 
@kirby@lab.nyanide.com I heard that Pleroma's postgres db is usually larger than Misskey's :satrithink:
Like Pleroma is already using 1gb for just a few users while a Misskey that has plenty of users (like 50) is at around 400-500mb :sagume_think:
is happening
@i
Good point!
Mine was that throwing JSON in original form right into DB has some advantages too. TBH I was surprised too when I discovered that, but now I thing I can see some reasoning behind it.
As for scalability — you're right of course, but ActivityPub itself isn't very scalable — if I designed Fedi from the ground up, I'd probably do a lot of things differently and implement it as sort of bus with guaranteed delivery and all that — probably won't work well in real world, with the way some run their instances, they might go offline all of a sudden, then come back up unpredictably. Sometimes I'm surprised Fedi works at all 😂
I'm not well-versed in how Misskey works, but it seems to me it tries to work this around as much as it's possible — and that makes the queues grow quite big and weaker hardware starts choking on it.
@kirby @mima @m0xee
Good point!
Mine was that throwing JSON in original form right into DB has some advantages too. TBH I was surprised too when I discovered that, but now I thing I can see some reasoning behind it.
As for scalability — you're right of course, but ActivityPub itself isn't very scalable — if I designed Fedi from the ground up, I'd probably do a lot of things differently and implement it as sort of bus with guaranteed delivery and all that — probably won't work well in real world, with the way some run their instances, they might go offline all of a sudden, then come back up unpredictably. Sometimes I'm surprised Fedi works at all 😂
I'm not well-versed in how Misskey works, but it seems to me it tries to work this around as much as it's possible — and that makes the queues grow quite big and weaker hardware starts choking on it.
@kirby @mima @m0xee
@m0xee@social.librem.one @i@declin.eu @kirby@lab.nyanide.com @mima@makai.chaotic.ninja yeah but that's because it's written in j*vascript 
so people have to keep buying ever bigger servers with ever bigger drives, instead of being able to split up the responsibilities in any way that's not handling user uploads
or continuously pruning the database, which means facing hours of performance issues while postgresql repacks all the messy data/indexes all over again