@waltercool @shebang @hfaust Well, the instance that I'm on is technically using a Mastodon fork.

https://github.com/matrix07012/mastodo

@xianc78 @waltercool @hfaust I wonder if they fixed the 50k emoji size limit on that fork.

Back when I ran a masto server I ended up doing a bunch of that shit, then later ran Mewmew's fork.

@shebang
I increased it to 1MB although the UI still says 50kb
@xianc78 @waltercool @hfaust

@xianc78 @shebang @hfaust Isn’t that just a fork made by @matrix for his/her own instance?

I can clearly see himself/herself as the only maintainer and code supplier.

This kind of acts have only two options:

You keep an old version of Mastodon for your own needs.
Your job will be basically merging all Mastodon code into your own fork.

So no real gain.

@waltercool
Yeah, that's essentially what I'm doing. I forked a fork mastodon.host used and I'm basically just maintaining it by merging upstream Mastodon and adding a few tiny tweaks.
@shebang @xianc78 @hfaust

@matrix @shebang @xianc78 @hfaust Aren’t you afraid of being very de-synchronized with main branch?

That’s not an easy job.

But uncool for Mastodon for keeping so many settings hardcoded. Pleroma does a better job on that.

In other hand, Pleroma single-thread architecture is a waste of CPU resources, specially about queue management. Sidekiq is far more powerful in any sense.

@waltercool @matrix @hfaust @shebang @xianc78 pleroma is single threaded? i thought erlang could thread.

@icedquinn @shebang @matrix @xianc78 @hfaust I only see one process playing along with my CPU, so no forking.

Mastodon delegates any parallel task to Sidekiq which spawns a dedicated queue and workers for real multi-thread management, but this is not cheap in resources, and that’s fine.

@icedquinn @shebang @matrix @xianc78 @waltercool @hfaust erlang does thread, extremely so that vps' with one thread need busywaiting turned off so it doesn't explode.

If anything pleroma's biggest fault is that it can't keep up with postgresql, two small cores will handle an instance right up until you can't write to disk fast enough, then even an overkill ryzen won't do shit, as poa.st has run into.

You can distribute the beamVM with little issues, you can't distribute pleroma's database at all currently.

@pleb have you tried that clustered postgres that microsoft bought out recently

@hfaust @matrix @shebang @waltercool @xianc78

@icedquinn @shebang @matrix @xianc78 @waltercool @hfaust no, there's ongoing work, by guess who, to find that sorta stuff out
https://git.pleroma.social/pleroma/pleroma/-/merge_requests/3585

@pleb i've never tried citus. i did try cockroach but it wasn't as drop-in compatible as suggested.

@hfaust @matrix @shebang @waltercool @xianc78

@waltercool
Yes I'm, but so far I'm managing it without issues. For that reason I don't make any big fundamental changes that could become an issue when merging.
@shebang @xianc78 @hfaust