@Sandra@idiomdrottning.org recently wrote Usenet had to die, arguing that Usenet was wasteful, everybody carrying a copy of everybody else, including spam and binaries. I agree with all the points made. The idea that there is one global network of groups, with everybody carrying all the groups and everybody being allowed to post led to the downfall of Usenet.
Having recently hooked my laptop and server server to a handful of other servers using UUCP, NNCP and NNTP for file copies, mail and news, I just want to emphasize that the same technology can be used to build a different net.
The spam problem arises when two systems "trust" each other and peer messages but one of the systems doesn't act on abuse and moderation reports. Perhaps Usenet generated too much of a workload for system administrators so they stopped following up on those reports. Perhaps some servers had become too big to ban.
I think we have the exact same problems with fedi. Some instances seem too big to defederate from.
Unfortunately, individual blocking doesn't work. If my posts end up on an unmoderated server because somebody I like is following me, the post shows up in the federated feed on that server and all the other users that I don't like can comment and those comments are sent back to me. I can block them all individually or block the large instance including my small number of friends. That instance has become too big to ban.
To come back to net news: From my point of view, the protocols are not to blame for the downfall of Usenet. It was the lack of moderation that killed Usenet. It was the ease of setting up web forums and being able to moderate them that made Usenet irrelevant. In the end, to all these people federation was not as important as moderation.
Administrators were unwilling to drop the groups that were overrun with people that needed moderation. This is a problem on all platforms with groups, e.g. Reddit.
Administrators were unwilling to ban the instances that didn't moderate their users. This is a problem on all federated platforms, e.g. Mail.
Administrators were unwilling to create new groups for people in their niches. At least nobody seems to be making that mistake again. IRC, Reddit, Discord – they all demonstrated that ad-hoc group formation and self-moderation was viable.
This is why I think moderation, group forming and selectivity of federation are the most important quality of a platform.
We can have a small network of servers offering mail and news exchange, like in the old days. The problem is unmoderated growth.
#Social Media
---
⁂
Ah, that's not a super new post, it's from September 2023. I dug it out again since we were talking about Illuminant, which seems awesome. It uses hashtags only as post keywords and instead, if I understand things correctly, relies on FEP-1b12 to provide group support.
I think we have the exact same problems with fedi. Some instances
seem too big to defederate from.
It's not quite the same thing.
Yes, it's a bad thing that I've being hesitant to defederate from m.soc even though I get more spam from it over Fedi than I've been getting from email fro the past several years.
But I'm still only getting individual posts from m.soc or any other server. (Which is also why hashtags don't work the way people think.)
On Usenet, the typical behavior would be to carry the whole thing. Currently, I get dozens of spam messages from m.soc. If I were to mirror the whole thing (which is possible with AP relays, and is what would be necessary for hashtags to work like channels, me mirroring the entirety of all the non-banned servers), I'd have to deal with thousands of messages. My server would croak.
Unfortunately, individual blocking doesn’t work. If my posts end up
on an unmoderated server because somebody I like is following me,
the post shows up in the federated feed on that server and all the
other users that I don’t like can comment and those comments are
sent back to me.
A problem. But again something that was way worse with usenet because on there, not only would I mirror or moderate abusive replies to me. I had to mirror abusive replies to *everybody*.
BTW, I'm, for the most part, glad that the servers-that-can-comment aren't allowlisted, that it works like email, that I can get comments from anyone. The downside of that is having to ban ban ban like crazy, but if it didn't work like that I'm not sure it even would be federated.
I can block them all individually or block the large instance
including my small number of friends. That instance has become too
big to ban.
A problem. But on Usenet, *every* instance was too big to ban since every instance was expected to carry everything.
Individual blocks is one way around it. That's why I, a small instance op, while I've banned many small mainly-abuse instances, my need to ban the big sloppily moderated silos like m.soc or threads is a li'l lower since I don't have hundreds of users to protect, which a midsize instance would have. I can absolutely see why midsize "safe space" instances would ban m.soc for example.
Also I "silence" every instance since I don't have a "whole known network" tab.
Killfiles was a user-level way to mute users on newsgroups. So there was *some* mitigation, but Fedi additionally has several other moderation tools like instance bans or silencing.
More fundamentally: ActivityPub is about sending individual posts. Usenet was about mirroring entire servers. Those two are not the same.
To come back to net news: From my point of view, the protocols are
not to blame for the downfall of Usenet.
The context that I wrote that Usenet post last September was to rant about hashtags. (A topic that I'm gonna return to every now and then until the group/tag situation gets a li'l better, which might require new UX affordances or perhaps only education.)
That problem, summarized:
And that's absolutely a protocol level problem!
A protocol level problem that Usenet had even without hashtags. (Well, there was one hashtag: Kibo.)
I've said a couple of times that servers mirror each other and for the sticklers in the audience I've got to add that not literally *every* server carried *every* group. For example, some didn't carry Anarchists Lunatics Terrorists a.k.a. the `alt` hierarchy. But that's the fundamental intended flow. Group mirroring.
So in that way, Usenet is unlike mailing lists, where individual messages are sent to individual users. FEP-1b12 group servers on Fedi (like PieFed and MBin) are awesome because they get the best of both worlds of the "only send the relevant posts" nature of mailing lists, and the "we're hosting this post anyway so we only need to store it once in the SQL" nature of Usenet. As long as one user on a Lemmy server is subscribed to `@garderning@foo.bar`, like an email user might be subscribed to a mailing list, it might-as-well display a copy of that group since it's getting those messages anyway. It's not perfect and we're still paying a hefty "decentralization tax", but it's more robust than mail and news.
It was the lack of moderation that killed Usenet. It was the ease of
setting up web forums and being able to moderate them that made
Usenet irrelevant.
Naw, Usenet kept puttering along after that, for a while, until the cost (and legal danger) of mirroring huge binaries archives which was the nail of nails. And that was a protocol level issue.
Fedi absolutely has a moderation problem. That is being addressed although it seems to me that there are still unsolved problems.
Usenet had a much bigger moderation problem and much of the reason for that was on the protocol level. We have instances and instance bans to help moderate clusters of users where each instance op helps moderate the users on their instance. Moderation is happening on a user level which makes much more sense than on a group level. (Someone can be a saint on `alt.suicide.holiday` but a jerk on `alt.swedish.chef.bork.bork.bork` for example.)
But Usenet also has a scalability problem. Resource use. Better moderation wouldn't be nearly enough to fix that.
Administrators were unwilling to ban the instances that didn’t
moderate their users. This is a problem on all federated platforms,
e.g. Mail.
Email can't moderate their users since email isn't supposed to be world-readable. (With exceptions like mailing lists; we can ban someone who sends abusive messages to those.) But even so, instance banning is very common on email.
We can have a small network of servers offering mail and news
exchange, like in the old days. The problem is unmoderated growth.
Last time I talked about this I got the world's most stubborn reply guy on my neck who (since he was on a huge instance) could use hashtags to find at least some amount of interesting posts and was resistant to change. Since I mentioned scalability issues and how my server would croak if it were to try to mirror even *one* of the big instances, he started going on and on about "scalability is only for capitalists".
Yet another case of Smolnet Wishful Thinking Syndrome. The kind of thinking that hates Eternal September more than anything but keeps designing protocols that lead to more Eternal Septembers since they're about everyone-seeing-everyone and then they resort to technical gatekeeping or even bullying to keep people out. Not into it.
My criticism of Usenet's protocol here has nothing to do with capitalist scaling, which is about "throwing computers at it". Sure, that'd be one solution to the protocol problems inherent to Usenet and relay-laden-ActivityPub.
Real Smolnet scaling is about designing protocols that are about connecting people to people (and to topic groups or other communities), and in a non-wasteful, mottainai way create small overlapping villages instead of one big shout-loud-and-go-viral channel.
There are two kinds of protocols. Those where everyone must see everything (like Antenna and CAPCOM, and Twitter), and those where that's not the case (like ActivityPub and email).
Here are three network layouts:
First we have a centralized network. One hub server and then every client is a spoke connected to that center core. The most efficient way to organize an "everyone-must-see-everything" protocol, but vulnerable to disasters like meteor strikes, nuclear explosions, billionaire takeovers, or earthquakes.
Then we have a decentralized network. Several hubs connected to each other, mirroring each other. Usenet worked like that. The hubs need to be just as beefy as the ones in a centralized network. This has some advantages like redundancy and robustnesss, and it's more politically appealing to the anarchists that built the internet. (Not sure why since there's still two tiers and a mod group that rules everything for everyone.) It can be wasteful. It can handle everyone-must-see-everything protocols. It can also really shine when everyone *doesn't* need to see everything, like IRC which is super efficient. IRC is a brilliant protocol design that, like a mailing list, only sends the messages to a server that that server needs.
Finally we have a distributed network. No hubs. It looks more like a mesh where every machine is connected to only a handful of "nearby" machines. This type of network is an absolute disaster for the everyone-must-see-everything scenario! Redundant to the point of wastefulness. That's a future we must fight tooth and nail! And that's exactly where the relay-centered version of ActivityPub was headed, and that's what would be "needed" for hashtags-as-groups-and-discovery to "work".
Here's where we the grassroots federalized ancom degrowth community need to check ourselves before we wreck ourselves because even though energy and e-waste is underpriced, a capitalist for-profit silo company would never in a million years set up a system like this because it's too wasteful ever for them, while we're liable to do so since each instance op only sees their *own* costs, not the cost for the network as a whole. The megacorps are OK with centralization because for them, the power consolidation in the first two network types is a positive, while for us it's a huge negative so we're drawn to the third type, so we need to be really careful and really responsible here.
energy and e-waste is underpriced
But as bad as this layout is for everyone-must-see-everything, it's really awesome for protocols where everyone *doesn't* need to see everything. It's economical, robust, cozy. ActivityPub (without the relentless mirroring) is a good fit here. I can follow a handful of users and be connected to their instances and only their posts are sent to my instance, I don't mirror the entire server. It's actually pretty awesome! It's good for moderation too because I only need to moderate my "neighborhood". Yeah, a "neighborhood" is the perfect analogy for this; neighborhoods that can and overlap but are still small and managed. A system set up like that can remain smol and cozy even as it invites everyone on Earth. It's great. ActivityPub is awesome.
---
⁂
ActivityPub hashtags are translated to Keywords: and vice versa in Illuminant. So far I haven't had the need to use hashtags/keywords for more than showing/writing them in Keywords: - I did have a plan to have groups like fediverse.keyword.music, fediverse.keyword.askfedi etc. etc. but I haven't implemented it.
I'm sorry to say that I do not know what FEP-1b12 is, so I don't think Illuminant implements any of the group support defined in that.
Since Illuminant is very much driven by what I personally want and feel like implementing/have the energy to implement, it has a kind of skewed subset of what I guess is expected from a fediverse server/user interface by most...
While it is fun for me to hear somebody call Illuminant awesome, I must say that I'm pretty sure there is only one person in the universe who actually runs it. Being proven wrong would be awesome, though.
