jon is a user on toot.cat. You can follow them or interact with them if you have an account anywhere in the fediverse. If you don't, you can sign up here.

@Elizafox not sure it even is really

it is one of about a half dozen things on my list tho

i have a way to do it also just... requires a little more plumbing than most of my fixes and i'm afraid of how it might perform

@bea @Elizafox

Using a SAT-based algorithm to do the necessary filtering would be more efficient than current implementation, actually.

@bea @Elizafox

the current post filtering implementation that is basically worthless because it doesn't let you set any actual rules :(

@bea

the present user/media blocking implementation

@bea

no, right now, when a message is received by mastodon, it pulls in what local doppelganger user would have created it (which is slow), and pulls in various objects for the local opinion of the remote server's policy to determine if media should be blocked or the message entirely.

instead, this needs to be controlled by a SAT with the local server policy stored as constraints in the SAT model. then you solve the SAT model with each message as the input.

1/2

@bea

such an approach is way more flexible than what mastodon has right now (user/domain blocks, media blocks), because it is a unified filtering model. implementing additional filtering constraints becomes virtually cost-free.

the SAT model (at server scope, and user scope) can be cached to make it very inexpensive, verses the current approach.

@bea i would code it at this point but i don't know a damned thing about ruby and don't have time to learn :(

but a SAT model could allow any attribute of a message to be filtered. :)

@kaniini eh well

i'm quite familiar with ruby, rails, sidekiq, and all the back end filtering code and this feature is important to me...

@kaniini @bea go for it!

Now would it be easy to keep it compatible with the rest of Mastodon?

@Balor @kaniini i don't think such a modification would rely on changes to other instances to function

@kaniini @bea

Right, but since it's a change of architecture it would add work to keep with the upstream, to repurpose patches to the new architecture.

@Balor @kaniini ideally it goes to upstream

if it doesn't ... at least it would be maintained in the glitchsoc fork for the foreseeable future

jon @jdp23

@cyrinsong this relates to the "value of a fork" discussion we had at

· Web · 0 · 1