peach eating vagus nerve cultist of the house of tool ape


DecayWTF
@DecayWTF

The Internet is not designed or set up to support a useful social network.

Consider the origins, what the Internet is actually designed for: Sites connecting to sites. Historically you would have a site that was basically one computer, maybe a PDP-11 for a smaller site up to a CDC 6600 supercomputer - the early arpanet maps were basically shitloads of DEC machines and a smaller number of Control Data and UNIVACs with a handful of IBMs for good measure - with a bunch of Honeywell 316s functioning as gateways and routers. Each site was its own network, its own community. Everyone worked from terminals on the big box. And that's what the Internet was designed for. Email is an obvious example of how that works in practice, and in that environment all this shit makes sense. Your local social network was mainly the people on the same site and you might just talk to them via a local bulletin board or in class or at the coffee shop.

Look at the Fediverse, it is designed on exactly those lines. Each Masto instance models an arpanet site. But until and unless there's more physical and meaningful community both per site and between sites - until there is some cost to defederation in the same way there would have been if, say, Stanford decided they were mad at BBN and didn't want to talk to them, which would have been horrendously damaging to both sites - trying to build new services along this model isn't going to work. Conversely, a pure peer-to-peer system like the telephone network isn't how the Internet is structured. Building something like that without any server or centralized infrastructure is basically infeasible because it violates a ton of basic assumptions the Internet is built around.

Forums were a good compromise. So are sites like Cohost. The Fediverse in a different and more well-thought-out form could be a good way of bridging that history with what people want but only if we as a civilization were willing to change the way we think about these things.


pnictogen-wing
@pnictogen-wing

...isn't this a problem with "The Web" and not "The Internet"?

remember NNTP? that was a better scheme for hosting communities than web pages, surely?

~Chara


DecayWTF
@DecayWTF

It's very much a problem with the Internet and not the Web alone. Usenet did face a lot of the same problems and there were many familiar consequences:

  • When Usenet moved away from UUCP to TCP/IP, a lot of small sites disappeared and major backbones became larger and gained more control because fatter pipes and the niceties of Internet routing meant that they could and it made Usenet run better.
  • Governance for the whole network was effectively in the hands of the admins if the backbone sites.
  • Individual users were at the mercy of their site admins; usually this was fine, sometimes it meant people could lose their news access because the sysadmin was in a bad mood.
  • Which newsgroups you could access was entirely at the mercy of the admins, there were plenty of sites that didn't carry alt.* and some groups in the more "reputable" hierarchies regularly got censored by assholes (soc.motss being a regular victim)
  • Spam sites were common and had to be handled with a variety of methods, some heavy handed.

A lot of very familiar issues the Fediverse is facing right now.

Edit: oh yeah don't get me wrong piggybacking on HTTP for distributed L7 systems sucks, it's just not the proximate cause of these issues.


You must log in to comment.

in reply to @DecayWTF's post:

we've had a peer to peer reading group going for awhile at Queerius Labs (or, rather, they do and I attend as a visitor) via discord

and the biggest commonality is peer to peer apps going "we need to reinvent the internet for it to be peer to peer, so instead we just do peer to peer routing over the normal internet because that doesn't require an infrastructure revolution"

and I do wonder if 5g is going to bring that one good thing into the world, might make up for the gouging. Especially if ad hoc 5g becomes a well-promoted standard, which the telcos will fight tooth and nail but would make a good grass roots movement.

Yeah, I messed with mesh networking software like ten years ago that did something similar. It was hugely deficient for a lot of reasons but conceptually it was interesting and I think that kind of distributed cache is a perfectly cromulent solution

We absolutely are and do constantly. The idea that the current state of society is some kind of natural occurrence rather than something that has been very carefully and deliberately constructed and requires constant violent maintenance to keep from falling apart is one is the biggest liberal myths.

until there is some cost to defederation [...] trying to build new services along this model isn't going to work.

I'm curious if you'd be willing to elaborate here - isn't it the case that the cost in the current ActivityPub model is about the same as in the ARPANET example - namely, mutual loss of access? What makes the Stanford and BBN machines different from two Mastodon instances, other than perhaps number of users and importance of those users' use cases?

Well it's that specifically, this is why I called out the fundamental difference between a typical mainframe site and eg a mastodon server. There's material consequences if Stanford people can't contact their vendors/fellow workers at BBN. If cheesewheel.fuck defederates gts.somedude.io because the latter got added to fediblock out of spite, there's no material cost to them, they may not even realize they did it. In the former case, at some point the site admin is responsible to their users. In the latter, in the median case the server admin is usually not responsible to anyone in any meaningful way. This is a social problem.

Gotchya, that totally makes sense to me. In the cases you're describing, you're absolutely right, but I have to say that my experience has been that good site admins and mods are pretty accountable to their users, and "got added to fediblock out of spite" is not typically considered an acceptable reason for defederation on a responsibly moderated server.

That is to say: I agree that it's a social problem, but I'm not sure that it's completely insoluble, nor indeed is it currently unsolved.

That's a lot of couching and ultimately all of this is dependent on the good will of the site admins because some of them may have good intentions of accountability but they're not materially responsible to the users. This is my point, the structure of the Internet and so of most or all services that run on it is derived from a set of conditions that no longer exist. What happens when Facebook federates and by virtue of the size of the user base becomes the de facto governance of the entire network?

ultimately all of this is dependent on the good will of the site admins because [...] they're not materially responsible to the users.

I mean, they aren't by default, just like the de-facto leaders of any social group aren't materially responsible to the members of that group. I think that has to be handled like any other situation where some person or group has power over others, right? Mitigations, governance structures, and so forth. There are several extant co-ops that run Mastodon servers, for instance, which really helps get around this problem.

Regarding Facebook, I think that's quite a leap; just being big doesn't guarantee control, as we've seen with Pawoo, who are huge, but are generally not able to access much of the Western fediverse due to lax rules around sexual images depicting minors.

Again, none of this is a meaningful counterargument. When I say, "the structure of the Internet is inimical to what people generally want out of their social communications", saying "but people are managing to make it work anyway" is not helpful. And comparing pawoo, which doesn't even mainly cater to English speakers, to Facebook is pretty silly. It's not a leap at all to say large sites control and eventually break federated systems, is literally exactly what happened with email, which has much less reliance on a fully-connected graph than the Fediverse

Ah, gotchya! I think I understand better what you're saying in that context; in particular, I wonder how much portable identity (not something the Fediverse, in any of its incarnations, has ever been able to do) would be helpful with this? That way, sites (in the sense you use it in the OP) could be entirely about infrastructure, rather than identity.

I'm also really curious - and I know this is going beyond your OP - what you think a good network structure for underpinning a social medium would be? Like, if we had an opportunity to redo the whole thing from scratch with people at the center?

Portable identity is an interesting idea, yeah, it would uncouple "account" in some sense from the sites themselves which would alleviate the problem of people being tied to specific instances of a network; IRC services were sort of a limited implementation that let lRC stay "open" within a given network but people could still claim their name.

As for network construction... I'm not really sure, but probably you'd need again some kind of way to allow people to be uncoupled from their physical or logical location in the network; a Person as a first-class routable entity. A more limited mechanism could probably be achieved once ipv6 deployment is complete and there's enough addresses to assign every person, place and thing in the universe a globally-unique IP and then some mechanism to update routing tables so that your IP could go with you wherever you were. This has other serious challenges of course, like privacy...

in reply to @pnictogen-wing's post:

same. I wouldn't trust any public servers. there might be some universe of private ones out there (and I'll bet there's some really nasty pits).

but we've been thinking of looking at the protocol itself? maybe some use can be gotten from NNTP from the ground up. it's ambitious I know

~Chara

in reply to @DecayWTF's post: