[bitwize]

One of the points that Aurynn Shaw makes is that federation hampers the addition of protocol features because new features have to be agreed upon by all participants in the federated network. This adds friction to the evolution of protocols, and results in the phenomenon wherein Slack runs halfway around the world while IRC is still putting on its shoes.

She also makes some excellent points about the social characteristics of federated networks. In a federated environment, each node sets their own policy, which meant there was no single point of responsibility for onboarding new users, setting standards of behavior, or filtering out trolling and harassment. Often, as on USENET and IRC, it is the user's own responsibility to filter out content and users they don't want to see -- and there was no authoritative source for new users to determine who/what should be filtered. Some USENET groups were moderated, and on IRC, channel ops can monitor and ban users for in-channel behavior, but if someone is harassing you in PRIVMSG there's often nothing you can do -- and channel bans can be circumvented and enforcement bots subverted fairly easily. And no one takes responsibility to communicate which instance of a federated network to join if they don't want to see particular kinds of content.

So federated networks quickly become cesspools of the worst forms of communication because they're optimized to promote all forms of communication -- "freedom of speech at all costs" as Aurynn says (and, as she points out, is actually a US-chauvinistic perspective on speech and runs contrary to the laws on speech even in most democratic countries -- hate speech being an offense is the norm). This tends to make them grognard-friendly, but hostile to new users and to users of marginalized communities, as well as potentially illegal to participate in in countries not called the USA. And that was accepted in the 90s internet because that's how the 90s internet was. But standards have changed and this is no longer acceptable. "Me too" has gone from the mark of a clueless n00b to a rallying cry against harassment. And people like Aurynn Shaw and Coraline Ada Ehmke have been leading the way in terms of calling out and removing the negativity, exclusion, and sometimes outright hate, from open source development communities with things like Coraline's code of conduct and Aurynn's efforts to highlight contempt culture -- the "PHP sucks" and "Micro$oft sucks" culture that prevailed in technical circles in the 90s and early 00s whose toxicity is something we still deal with today.

Times have changed since federation came out and was promoted as a wonderful thing. Back in the 90s, we thought that building the technology itself was sufficient to change the world for the better. Today we understand better the social costs that mentality has unleashed. We optimize for creating safe, welcoming communities and promoting voices that are usually silenced, rather than allowing everyone to communicate anything at any time. Unfortunately, federated technologies as we understand them today still come from that 90s mentality, and without broader conversations about the social impacts -- as well as establishing some sort of standards for mitigating those impacts -- it's simply better to not federate. Slack and Discord are easier to get started with, offer more features, and promote a safer and more welcoming environment than does IRC.

[progval]

> but if someone is harassing you in PRIVMSG there's often nothing you can do

The large majority of IRC networks are centrally managed these days, so you can talk to the network operators so they ban the person from the network.

> channel bans can be circumvented and enforcement bots subverted fairly easily

Disallow unregistered users on your channel, and it becomes as hard to circumvent as registering on any other platform; assuming the network uses standard blocklists like DroneBL

[caslon]

I agree that federation is bad, but "Just use [two awful proprietary services, one owned partially by the unethical Chinese conglomerate Tencent and the other owned by a terrible American company]" isn't a good alternative.

[qu4z-2]

If the problem with federated networks is that they can't be globally censored to your (and Coraline Ehmke's) standards, I think I will continue to use them.

[account42]

I'd take all the spam, trolling and harassment in the world over letting some would-be censors decide what I can and cannot see, or say to others who want to hear it.

We should all be thankful that so far neither corporate nor authoritarian have managed to completely kill the biggest federated service we have: EMail.

That federation makes authoritarian goals harder is a problem for authoritarians, not a problem with federation.

> the "PHP sucks" and "Micro$oft sucks" culture that prevailed in technical circles in the 90s and early 00s whose toxicity is something we still deal with today.

Both of these have been righly criticized for many reasons and trying to frame that as some kind of bad thing does not invalidate the criticism.

[Klonoar]

>One of the points that Aurynn Shaw makes is that federation hampers the addition of protocol features because new features have to be agreed upon by all participants in the federated network. This adds friction to the evolution of protocols, and results in the phenomenon wherein Slack runs halfway around the world while IRC is still putting on its shoes.

This is effectively the point Moxie was making with Signal ("the ecosystem is moving").

I don't think that this means decentralized services (IRC or Matrix or insert-your-thing-here) shouldn't exist, but it's pretty much proven at this point that centralized services will always gain the lions share.

[sweetbitter]

> One of the points that Aurynn Shaw makes is that federation hampers the addition of protocol features because new features have to be agreed upon by all participants in the federated network. This adds friction to the evolution of protocols, and results in the phenomenon wherein Slack runs halfway around the world while IRC is still putting on its shoes.

This is true in the case of a fragmented specification, however there is a neat middle-ground that has been observed recently in the case of the Matrix specification, in which a single team has been setting the spec that many clients implement much quicker than would occur in the environment of IRC. The spec's movement will eventually become much more solid than its current liquid-ish state, but by that point it, the specification itself, the set of implementations, and the collective network effect will be quite valuable.

> ... each node sets their own policy, which meant there was no single point of responsibility for onboarding new users, setting standards of behavior, or filtering out trolling and harassment. . . .no authoritative source for new users to determine who/what should be filtered. . . . behavior, but if someone is harassing you in PRIVMSG there's often nothing you can do -- and channel bans can be circumvented and enforcement bots subverted fairly easily. . . .

This is not an issue, as long as the user is told the ramifications of federation, that curation of what they can see is done either by themselves or by the particular server they use. Not to mention that it is just as difficult to subvert bans on a network like Matrix, or a registered-user-only chan on IRC, as it is on the likes of Discord.

> So federated networks quickly become cesspools of the worst forms of communication because they're optimized to promote all forms of communication -- "freedom of speech at all costs" as [. . .] This tends to make them grognard-friendly, but hostile to new users and to users of marginalized communities, as well as potentially illegal to participate in in countries not called the USA.

If some speech is illegal to in your country and you also do not think it is moral, you should not speak it. Otherwise, why do we need some central authority to micromanage us and determine what types of speech are to be blacklisted? Even pretending that such a thing could be accomplished in a scalable way, and pretending that the codification of 'illegal numbers' is a good idea, why is it necessary to introduce such a vulnerability into communication software? The admins of a server may curate as they wish and users may choose which communities they wish to engage in by virtue of their choice of server and subscriptions to preferred data.

> Today we understand better the social costs that mentality has unleashed. We optimize for creating safe, welcoming communities and promoting voices that are usually silenced, rather than allowing everyone to communicate anything at any time.

If anything, I wager it is the opposite. The current model in which the position of censor is centrally held is far more prone to abuse than its predecessor, considering the impossibility of that authority to reconcile with the unlimited number of complaints various groups will have with its decisions. Things are far more toxic and unwelcoming today under this model than they have ever been before, with more innocent voices silenced than ever were before, precisely because the typical user no longer controls their subjective flow of information. People should have the capacity to curate for themselves, to control the machine and not be controlled by it and the whims of vendors, to control both what they wish to say and what they wish to hear. If that were still the case today, we would never have gotten ourselves into this mess.