>but you need it to be architectured in such a way that no organization can capture the network effect in order to set up a choke point.
How is that not the case now?
>You need all moderation to be applied on the client, or you'll have large servers doing things like banning everyone from new/small independent servers by default so that people have to sign up with them instead.
I suppose. There are ActivityPub "clients" which act as interfaces that allow the former and act as agents for a single user interacting with other ActivityPub instances. which, I'd expect can take us most of the way you say we should go.
I haven't seen the latter, as there's really no incentive to do so. Meta tried doing so by federating (one-way) with threads, but that failed miserably as the incentives are exactly the opposite in the Fediverse.
I suppose that incentives can change, although money is usually the driver for that and monetization isn't prioritized there.
>The protocol needs to make that impossible or the long-term consequences are predictable.
Impossible? Are you suggesting that since ActivityPub isn't perfect, it should be discarded?
ActivityPub is easily 75% of where you say we should go. Much farther along that line than anything else. But since it's not 100% it should be abandoned/ignored?
I'm not so sure about your "long-term consequences" being predictable. Threads tried to do so and failed miserably. In fact, the distributed model made sure that it would, even though the largest instances did acquiesce.
ActivityPub is the best you're going to get right now. and the best current option for distributed social media.
Don't let the perfect be the enemy of the good.
Edit: I want to clarify that I'm not trying to dunk on anyone here. Rather, I'm not understanding (whether that's my own obtuseness or something else) the argument being made against ActivityPub in the comment to which I'm replying. Is there some overarching principle or actual data which supports the idea that all social media is doomed to create dystopian landscapes? Or am I missing something else here?
The protocol allows servers, rather than users, to ban other servers. Servers should be only the dumbest of pipes.
> Are you suggesting that since ActivityPub isn't perfect, it should be discarded?
I'm saying that by the time something like this has billions of users the protocol is going to be a lot harder to change, so you should fix the problems without delay instead of waiting until after that happens and getting deja vu all over again.
> Threads tried to do so and failed miserably.
Threads tried to do that all at once.
The thing that should be in your threat model is Gmail and Chrome and old school Microsoft EEE. Somebody sets up a big service that initially doesn't try to screw everyone, so it becomes popular. Then once they've captured a majority of users, they start locking out smaller competitors.
The locking out of smaller competitors needs to be something that the protocol itself is designed to effectively resist.
>The protocol allows servers, rather than users, to ban other servers. Servers should be only the dumbest of pipes.
A fair point. A good fix for this is to have individual clients that can federate/post/receive/moderate/store content. IIUC, there is at least one client/server hybrid that does this. It's problematic for those who don't have the computing power and/or network bandwidth to run such a platform. But it's certainly something to work towards.
>> Are you suggesting that since ActivityPub isn't perfect, it should be discarded?
>I'm saying that by the time something like this has billions of users the protocol is going to be a lot harder to change, so you should fix the problems without delay instead of waiting until after that happens and getting deja vu all over again.
I'm still not seeing the "problems" with server usage you're referencing. Federation obviates the need for users to be on the same server and there's little, if any, monetary value in trying to create mega servers. Discoverability is definitely an issue, but (as you correctly point out) should be addressed. It is, however, a hard problem if we want to maintain decentralization.
>The thing that should be in your threat model is Gmail and Chrome and old school Microsoft EEE. Somebody sets up a big service that initially doesn't try to screw everyone, so it becomes popular. Then once they've captured a majority of users, they start locking out smaller competitors.
Given the landscape of the Fediverse, that seems incredibly unlikely. Perhaps I'm just pie in the sky on this, but those moving to ActivityPub platforms do so to get away from such folks.
Adding to that the ability to manage one's own content on one's own hardware with one's own tools, it seems to be a really unlikely issue.
Then again, I could absolutely be wrong. I hope not. That said, I'm sure that suggestions for changes along the lines you suggest to the ActivityPub protocol[0][1][2] as a hedge against making it fall into a series of corporate hell holes, as you put it, "impossible," would be appreciated.
How is that not the case now?
>You need all moderation to be applied on the client, or you'll have large servers doing things like banning everyone from new/small independent servers by default so that people have to sign up with them instead.
I suppose. There are ActivityPub "clients" which act as interfaces that allow the former and act as agents for a single user interacting with other ActivityPub instances. which, I'd expect can take us most of the way you say we should go.
I haven't seen the latter, as there's really no incentive to do so. Meta tried doing so by federating (one-way) with threads, but that failed miserably as the incentives are exactly the opposite in the Fediverse.
I suppose that incentives can change, although money is usually the driver for that and monetization isn't prioritized there.
>The protocol needs to make that impossible or the long-term consequences are predictable.
Impossible? Are you suggesting that since ActivityPub isn't perfect, it should be discarded?
ActivityPub is easily 75% of where you say we should go. Much farther along that line than anything else. But since it's not 100% it should be abandoned/ignored?
I'm not so sure about your "long-term consequences" being predictable. Threads tried to do so and failed miserably. In fact, the distributed model made sure that it would, even though the largest instances did acquiesce.
ActivityPub is the best you're going to get right now. and the best current option for distributed social media.
Don't let the perfect be the enemy of the good.
Edit: I want to clarify that I'm not trying to dunk on anyone here. Rather, I'm not understanding (whether that's my own obtuseness or something else) the argument being made against ActivityPub in the comment to which I'm replying. Is there some overarching principle or actual data which supports the idea that all social media is doomed to create dystopian landscapes? Or am I missing something else here?