deadsuperhero

I wrote a counter-point to this a while back: https://wedistribute.org/2024/05/forking-mastodon/

I'm not saying "don't do it", but realize that the amount of commitment required to make a hard fork even moderately successful is vast.

It's telling that the biggest project in the space is barely able to pay more than a handful of people to work on it, and it still develops at a snail's pace. Notably, those are the people who deeply understand the system and its internals. While it's not impossible, you have to be realistic about how much further a group can get when they don't have the insight or technical chops required to take development further.

Yes, the entire platform trains itself on posts within its platform to make algorithmic decisions and present it to users. Instead of likes or follows, you just have that.

Pretty much everybody.

They kind of fucked up everything in approaching this by not talking to the community and collecting feedback, making dumb assumptions in how the integration was supposed to work, leaking private posts, running everything through their AI system, and neglecting to represent the remote content as having came from anywhere else.

The other thing is that Maven's whole concept is training an AI over and over again on the platform's posts. Ostensibly, this could mean that a lot of Fediverse content ended up in the training data.

I'm like 99% sure that this is just WordPress with BuddyPress extensions. I could be wrong! But, if that's the case, it's likely all open source already. It might be interesting to see if they can get it working with the ActivityPub plugin for WordPress!

I mostly meant the old system where all options were just a bunch of floating neon text. It was...functional, I guess, but the new system is so much better from a controller perspective. It's not perfect yet, things like elevator buttons are still in a weird spot, but most things are a lot easier to deal with.

One edge case that I really want to see the team nail down is crash loops. In the EPTU build of 3.23, we noticed situations where recovery effectively acted like a rewind feature, but didn't actually prevent the cause of a crash from happening.

Having to experience a handful of the same crashes during a single play session is pretty painful.

Honestly? I'm loving it. The biggest improvement for me was getting rid of those awful PIT menus that were ugly and sometimes hard to use. The new system is way more usable, and I'm tweaking the mappings on my controller to see what feels the most usable.

The improvement to EVA is also phenomenally good. You move a bit faster, there's more precision, and traversal between EVA and ship is much smoother. As a salvager that gets in and out to scavenge cargo holds, this is a big deal to me.

The character customizer is also really fun to use, and feels pretty intuitive to use. There's still work to be done in explaining what all of these vertices do, but I think the customization is a lot more flexible.

Some pretty nasty bugs emerged in 3.23 and 3.23.1, but it seems like the team is making pretty good progress on improvements? So, there's that.

I can't tell whether this is serious or sarcastic 😅

As far as the "global square" part of the equation is concerned: yeah, you're right! A firehose of public statuses requires indexing to work, as a basic foundational premise.

However, there's nothing preventing someone from standing up a PDS, opting out of the firehose / big graph service, and instead leaning on federation between individual PDSes. I'm not saying it would necessarily be a common use-case, but it's definitely not impossible.

It's a different approach with different ideas. It uses open protocols, focuses on data and account portability, and incorporates peer-to-peer concepts in its architecture. The vision behind Bluesky is to build a global square with these concepts.

I definitely wish they would've extended ActivityPub and collaborated on the wider network, but I kind of understand wanting to start from scratch and not get involved with the cultural debt Mastodon brought to the network.

Misskey is a little bit odd, in the sense that there's constantly new forks in various stages of development. New forks emerge just as quickly as old ones die off.

It may be that the frontend and backend both being written in one language helps make the system easier to hack on. I can't say for sure. What's weird is that some of these forks go in really odd directions, like rewriting the whole backend in a different programming language.

The other thing is that, despite their proliferation, the effort is somewhat fragmented into all of these little projects. I'm not sure how viable any of these forks are in the long term.

Thank you for these insights!

Yeah, aside from developer muscle, an effort like this requires deep knowledge of the existing system. Or, failing that, a commitment to learning it.

It's also not something that can be done as a side project, if it hopes to compete with the main project to the point of replacing it. Something like that requires an ungodly amount of effort and dedication. Someone would have to commit years of their life to solely working on that.