pleroma.site

abunhdhappyhop abunhdhappy abunhdhop abunhd abunhdhappyhop abunhdhappy | @kaniini@pleroma.site

a friendly #collectivist ๐Ÿฐ

destroyer of bloat @ #pleroma, #pkgconf, #audacious
slayer of techbros
previously #alpinelinux core a few moons ago and #debian much longer ago

she/her

@flussence as i keep saying, automation is the only way to have any hope of staying ahead of the shitshow. yes, automation can go wrong, but automation can be a huge win when it's working properly. enabling that kind of automation was the whole point of MRF. too bad it wasn't fully realized.

@flussence it's fine, just keep blocking instances or something. since really that's the only way anyone can curate anything.

@flussence i hate to be that person, but i warned *all* of these fuckers. they don't care. all they care about is DAU metrics.

@deavmi @sjw @HarneyBA @dee @velartrill @boob

i'm mostly uncaring, tbh. i just don't wish to deal with it personally.

@alex @HarneyBA @boob @dee @sjw

Mastodon can do that on 500MB, if you know what you're doing.

@deavmi @sjw @dee @HarneyBA @velartrill @boob

no. misskey is in nodejs because that is what the author uses.

@alex @HarneyBA @boob @dee @sjw

CPU time is the most expensive part. Many things downgrade to hash scans and seq scans because the indices fail to keep up. This is why "VACUUM ANALYZE" is a common suggestion for "fixing" Pleroma performance issues (temporarily).

@alex @boob @HarneyBA @dee @sjw @velartrill

once a Pleroma instance starts scaling out, it is basically as much of a nightmare to manage as Mastodon.

managing pleroma.site is a nightmare, almost everything on the fediverse is actively federating with it.

when i handed it over to href, the database was almost 300GB.

whenever href updates the instance, it takes several days to run migrations.

it took several days to import too, when i handed it over.

@boob @deavmi @HarneyBA @dee @sjw @velartrill

Misskey looks like a very nice platform. I don't want to deal with managing a NodeJS service for my own personal use though.

@boob @deavmi @HarneyBA @dee @sjw @velartrill

they could have done that by simply prohibiting DMs from leaving DM scope. no need to add a new object type.

@boob @HarneyBA @dee @sjw @velartrill

well, you see it from the end-user perspective.

from the admin perspective, it is largely "why is postgresql pegging CPU again today? why are queries falling over again?"

and from the dev perspective, it is largely "maybe if we add emoji reactions, we can get users"

when i was running instances, i deployed a shared postgresql for them that had 256GB of RAM allocated for cache. that made Pleroma fast, but it's not realistic.

@velartrill @HarneyBA @boob @dee @sjw

that is literally what i am building. also, the ability to make channels and identity portable (kind of like fedi, but done right).

@sjw @velartrill @HarneyBA @boob @dee

there are ways to integrate IRC and push messaging. that's something the IRCv3 group has unfortunately not spent much time working on, but it's possible

@velartrill @HarneyBA @boob @dee @sjw

ironically, the thing i am working on uses IRC as the signalling, providing a discovery layer for web services (like WebRTC) on top for the things IRC is bad at.

@boob @HarneyBA @dee @sjw @velartrill

the problem is that the majority of dev effort is focused on new features, and not fixing bugs.

@boob @HarneyBA @dee @sjw @velartrill

which would you rather have:

- a chat feature

- actual bugs in Pleroma fixed (like the fact that Pleroma suddenly forgets when random people are following you, slow timeline building, slow tag fetching, ...)

@hj @HarneyBA @dee @sjw @velartrill

do yourself a favor and push against the mastodon E2EE feature then. message franking is a disaster.

@sjw @HarneyBA @dee @velartrill

that's kinda sorta what i am working on these days instead of Pleroma. but we are in very early days yet of that.

@sjw @HarneyBA @dee @velartrill

that would be difficult to achieve because of the way tags are indexed in Pleroma. "everything is JSONB" sounds great until you have to actually deal with stuff like tags.

they could split tags out into a separate ARRAY column, but that would involve a heavy migration.

@sjw @HarneyBA @dee @velartrill

we can see this line of thinking with the new ChatMessage feature. they have hyped it up, but the reality is, nobody cares because folks are using a dedicated platform outside fedi for their messaging needs.

ยป