A credible threat to (and from) commercial social network silos/2

Foreword

When I got started on this series, the discourse was focused on the imminent launch of what is now known as Threads, the microblogging service offered by Meta, the parent company that also owns Facebook and Instangram and has more recently decided to trademark common words to make it harder for us to speak without mentioning it.

The choice to focus on it was motivated by the threat that this new platform poses to the Fediverse via the Embrace, Extend, Extinguish (EEE) strategy that has been deployed with mixed success by tech giants in the last decades: in this sense, the threat posed by Threads comes with their declared intention (so far only partially realized) to federate via ActivityPub.

There was enough material about that already to convince me to postpone discussing its main commercial competitor, Bluesky (BS), and the threat it poses to the future of the open web by choosing to not be compatible with the ActivityPub network, while still presenting as a champion of decentralization (more on this later).

The benefit of having delayed writing the article is that I can go more in depth with specifics about it. The downside for you is that you have a lot to read (and for me a lot to write —took me over a week to finalize the first public draft).

Fighting fire with fire

In this sense, BS and Threads together are a two-pronged attack on the Fediverse: the former as the (“better”, for appropriate definition thereof) alternative for decentralization, and the latter as the EEE Trojan horse for the case when the Fediverse still wins (more on this later, too).

This shouldn't be read as some sort of “conspiracy” (with Jack Dorsey and Mark Zuckerberg meeting behind closed doors to plan together a combined attack), but rather as the predictable outcome of competitors finding themselves in front of a common enemy (the Fediverse) that is gaining credibility and popularity, and adopting different strategies to bring it down while trying to trip the other at the same time.

The difference in these strategies reflect also in the different approaches the platforms take on content distribution and —most importantly— moderation.

Meta has extensive understanding of the importance of moderation to keep (a certain type of) users happy, even when failing at implementing it properly (and I'm not even talking about their role in the Rohingya genocide in Myanmar here). They keep getting it wrong (can't get rid of your biases and sympathies after all, and throwing automation into the mix just makes everything worse), but at least they try.

The Twitter founder, on the other hand, has a more, shall we say, “TESCREAList” approach to moderation, or if you'd prefer a more libertoloid1 approach to it, that essentially boils down to «I want Nazis on my platform». (You can read the jwz jab about it here, but I recommend reading the more detailed and documented write-up by David Gerard to have a more complete picture of how deep the rabbit hole goes.)

Of course, there's only two types of people that are OK with Nazis on their platform: Nazi sympathisers, and people genuinely deluded that it's possible to have a constructive debate with a Nazi other than punching them in the face. But I repeat myself. (By the way, I actually disagree with Mike Masnick on that one: the Substack CEO is well aware of what he's turning his platform into; it's not by mistake, it's by design, which is why I recommend looking into something like Ghost to anyone looking for a publishing platform.)

Don't worry, though, having been founded specifically to allow Nazis to thrive in the same social space as you, with a data propagation design that is inherently unsafe, isn't the only thing that sets BS apart from the Fediverse: there's also the cryptocurrency grift. This has roots that trace back to when BS was still “incubating” at Twitter (from the choice of Jay Graber as CEO to the cooperation with the (at the time new) cyrptogrift Twitter team that brought us those those wonderful NFT profile pictures, so functional in helping us identify fools with a simple look at the shape of said profile picture).

If I were a BS user, I would give particular attention to everything they promise not to do in their announcement because that's exactly the path they'll take when they start to cash in for the investors' exit. Shall we be thankful that at least they haven't yet made any announcements about SALAMI (“AI”) yet? (And yes, more on this later too.)

Leaving moderation and exploitation aside, let's bring back into focus the different paths Threads and BS are taking for federation. I've already discussed at length about how Threads' (selective) “embrace” of ActivityPub can be weaponized against the Fediverse, and as promised I'll spend a few more words about it later. For now, though, let's see how BS approaches it differently, and what this means for the Fediverse.

BlueSky “decentralization” theory and practice

One of the purported purposes of BS is to test, validate and promote the use of the AT Protocol (nothing to do with modem commands), the underlying protocol designed by the BS developers to separate “speech” from “reach” —the separation behind the idea of «I want Nazis on my platform, but some people may not want to see them, and I want to spill both for cash» for which BS was created in the first place— on the same network.

I mentioned in the Foreword that BS has been created in response to the threat posed by the Fediverse. This is clear from the timeline. Dorsey announced the BS initiative in 2019. (I have no doubt that the idea came from engineers that truly believed in said decentralization as an essential tool for the open web, but what happened to it after it got in the hands of the leadership is an entirely different matter.)

At the time of the announcement, the Fediverse had already established itself through the complex network of protocols that the FLOSS community had experimented with and developed over the years, from the diaspora* project born ten years earlier to build a federated alternative to Facebook, to the family of protocols (StatusNet, OStatus, ActivityPump) that converged into ActivityPub with the W3C standardization process that ended in 2018, passing through platforms such as Friendica and its fork HubZilla designed with multi-protocol interoperability in mind, and arguably the first concrete examples of implementation of essential principles such as nomadic identity.

(I stress the latter in particular because one of the purported reasons why ActivityPub was snubbed by Twitter when evaluating the existing decentralized protocols was the absence of a portable identity —which however in itself is a pretty poor justification for developing an entirely new protocol, as shown for example by the proposed extension by Mike Macgirvin, the author of Friendica, HubZilla, Zot6 and several other projects that include efforts to bring self-sovereign identities to the Fediverse. The curious may find additional information in the documentation about federating nomadic identities of the streams repository. But doing that requires having interoperability as a priority, and an interest in promoting open standards, which really isn't what BS was ever about.)

By 2022, even before the sale of Twitter became final, the mindful people (yes, I include myself in the group) had already realized that the handover would be fatal to the platform and had started looking for alternatives. Saturated by the enshittification wave that was covering all corporate platforms, they realized that the only path forward was to stay as far as possible from the control of the giants of surveillance capitalism and found in the Fediverse the better (or more appropriately “least worst”) alternative. One of the key advantages it had over the competition was existing, and having several years of mixed track record showing (some of) both its strong and weak points (positive example: the isolation of the fascist network Gab; negative example: the Will Wheaton experience).

Meanwhile, in the three years since its start (yes, I'm including the time preluding to the 2021 foundation of the corporation that was spun off to handle the development) the BS initiative had little to show for itself (as Jack Dorsey himself said in early 2022, “It has been slow”) giving several wannabes (from Hive to the soon defunct Post.news, just to name the first that come to mind) the opportunity to spread chaos among those seeking for a new digital home in the preludes and even more so in the heat of the first large-scale Twitter exodus following the handover in late 2022, be it to cash in quickly from VCs taking the opportunity to milk the cow of the Twitter-disillusioned, or to divert attention from the Fediverse —leveraging its limitations (both real and FUD-fueled) and wads of cash to build something æsthetically attractive but with no concrete (or questionable when present) vision for the future and the platform sustainability

This («oops, we're losing face») put pressure on BS to “put something out there”, which turned out to be underwhelming, with not much to offer over the several Twitter “alternatives” and look-alikes that had popped up to cash in on the Twitter exodus other than the promise of “the new underpinning technology of the AT protocol”.

As it turns out, the purported decentralization theoretically made possible by ATproto is largely performative. This isn't just a coincidence due to BS being (essentially) the only significant implementor of the protocol, but is quite clearly a design decision. This is discussed in the aforementioned article about why Jack Dorsey ultimately dropped BS, but I also recommend reading @rysiek​@mstdn.social's write-up on the topic (out of date on some aspects, but still largely on point) and the more recent thread following @jonny​@neuromatch.social's comments on the possibility of alternate relays in the so-called “ATmosphere”.

This performative approach allows BS to reap all the benefits of centralization, while still occupying the decentralization mindspace —where that matters— to the point of confusing less knowledgeable people about whether or not BS is part of the Fediverse or not (it's not; the fact that a bridge service exists that allows posts from the Fediverse to be propagated to BS and conversely, while providing a potentially valuable service, helps muddy the waters in their favour —more on this later too!).

Of course, being still essentially a centralized service despite the decentralization cosplay, BS doesn't actually suffer from any of the issues that come from actual decentralization, such as the directory fragmentation, message propagation and indexing issues, or even just the “dreaded” need to choose an instance that has apparently (or allegedly) scared so many users off the Fediverse, while still allowing it to build “cred” for ATproto to be “federation done right” simply because its claims of federation are essentially untested.

As such, it has significant appeal for those who have gained awareness of the “limited timespan” (among several other issues) of corporate silos and somehow got wind that a decentralized, interconnected network closer in spirit to the “older” Internet, far from being a step back, would actually bring a breath of fresh air into their passion for the medium, but at the same are unaware of the existence of the Fediverse, or more often than not simply got scared (frequently before even trying) by its purported difficulties: it thus deludes them into having found an “escape hatch” via its pretense of decentralization, diverting their efforts towards rebuilding their presence in yet another corporate silo instead of an actually revolutionary (if at time painful) endeavour.

This combination of simplicity from its essentially centralized nature on one hand, and the paint of freshness and innovation from the “magic” protocol on the other, has been a nontrivial factor in its adoption during the most recent exodus from the social silo formerly known as Twitter and renamed by its new owner to that unoriginal (and pornographic-sounding) “X” Musk has been obsessed with since his online payment days that has led several connoisseurs to refer to it by the more apt Xitter moniker, pronounced with an initial “sh” sound to signify the critical difference in management and moderation style between the old and new platform.

But is there anything behind the cosplay?

A bridge too far

Ironically, basically the only thing that gives a semblance of credibility to the “decentralized” claim of BS is … the Fediverse. And even that is not only risible (metric-wise), but actually dangerous (for the Fediverse).

So where does this decentralization come from? By means of the aforementioned bridging service, Fediverse accounts can follow (some) BS accounts and vice versa, with consent given by following the “bridgehead” (endpoint) represented by the Fediverse (resp. BS) account representing the bridge itself on either network.

Aside from some technical limitations, the fact that the bridge allows accounts on either side of the bridge to see posts from the other side as coming from a “meaningful” user is an indication that some amount of decentralization is indeed possible with ATproto. As explained in the write-ups linked above (1, 2, 3), however, ultimately the control on what is and isn't accessible to the majority of users on the network remains in the hands of the corporation that controls the main node (see also the Entryway discussion in the BS technical documentation).

This means in particular that at any point in the future BS can introduce changes, or even simply decide to cut off “non-sanctioned” servers at no cost for themselves. (This is not a hypothetical: if you are hosting an ATproto PDS you must keep it up to date because BS introduces backwards-incompatible changes from time to time, and that cuts off your connection to the rest of the so-called ATmosphere.)

If this is reminiscent to the threat from Instagram's Threads I discussed in the previous installment, it is not by chance, because the leverage these juggernauts have always comes from the same source: user count. (And that's without even considering that BS also has complete control of the protocol itself, and obviously of the reference implementation which is, AFAIK, what basically everybody uses, not to talk about the “distributed” identity provider all BS accounts use.)

We can look at the number game by comparison with the Fediverse. (Note that the following statistics do not account for the contribution from Threads, and not just because getting metrics about it is all but impossible —the best I could do was discovering that around 2K federated Threads accounts overall are visible from my Fediverse accounts on instances that didn't join the FediPact. I don't have an account on mastodon.social, though, which is probably the Fediverse instance that sees the most federated Threads accounts. Other estimates reportedly put it around 50K.)

According to FediDB, at the moment of writing, the total Fediverse user count is just above 11M, with a rather stable 1M “monthly active users” (MAU). The largest Fediverse instance, Mastodon's flagship, has around 2.2M users, with around 240K MAU. In other words, the largest Fediverse instance has around 20% of the total users, and 24% of the MAU. These are already considered excessive ratios within the Fediverse, but if for any reason mastodon.social decided to go its own way and cut off the connection to the other servers, they would cut off over three quarters of the Fediverse. The Fediverse would suffer a bit, but would continue to exist without them.

For comparison, according to Wikipedia, BS currently has over 13M users, with 6.8M monthly active (ratios not unusual for a new social network in its growth phase, especially after the mass migrations caused by Xitter's problems in Brazil first, and more recently by the Xitter changes about blocks and the future use of user content for training of SALAMI energy sinks). How much of that comes from “outside” BS?

For the answer, we look at the data collected by @mackuba​@martianbase.net, presented in some statistics and a directory with information about traffic from non-BS PDSes. As of today, for example, the stats show that over an approximate peak of 2M unique weekly posting BS users, there are than 6K that come from outside BS, and the vast majority (nearly 5K) of these come from the Fediverse “proper” (through the bridge, that also covers the Nostr network I may talk about at some other time, and “direct” bridging of websites), while less than 300 come data servers different from the bridge. From the directory we see that of the over 16K non-BS accounts visible on BS, less than 1K do not come from the bridge.

(Of note, the directory explicitly mentions that the list covers non-BS PDSes that are visible to BS itself. There may be more that are not federated. How many is not known, but I suspect that if their numbers weren't vanishingly small we would have heard about them. This is actually another aspect where the “ATmosphere” and the Fediverse differ: statistics about the Fediverse are much less accurate, and almost surely underestimated2 due to how vast and variegated the ecosystem is even across the federation (i.e. without considering isolated subnetworks like Gab) —some instance software, for example, doesn't even report total or active users— whereas the ATmosphere is essentially defined by “what's in BS' orbit”: rather than an atmosphere, it should be considered more akin to the solar system.)

This tells us a few things about BS for our “number games”:

1. around 80% of the non-BS traffic on BS comes from the Fediverse (through the bridge; this does not account for traffic coming from multi-protocol software such as Friendica instances with ATproto support, but again there's reason to believe that's vanishingly small);
2. around 96% of the non-BS accounts on BS come from the bridge;
3. the entire non-BS traffic on BS accounts for less than 3‰ (that's per mille, not percent) of the BS traffic

(Also fun fact: after each “Musk did something” peaks, the numbers of “native” BS posts shows a sharp decline, while that of bridged posts remains constant, when not showing straight out growth: i.e. relatively more traffic starts coming in through the bridge from the Fediverse. If the “native” BS traffic were to back to August level, the Fediverse contribution would jump up to around 1%.)

In addition to my initial statement above about the Fediverse being what really gives credibility to BS's decentralization claim, these numbers confirm, among other things, that BS could cut off all external PDSes and it would make barely a difference for them or their users, while everybody else in the “ATmosphere” would essentially remain isolated. This is not how decentralization works. (Compare and contrast with what would happen if the Fediverse juggernaut decided to isolate itself.)

Of course, this isn't going to happen anytime soon: as with any commercial platform in their growth phase, BS will do all they can to encourage developers and users to join in and build stuff that integrates with (i.e. depends on) their platform: they need to cultivate an “ecosystem” before moving on to the “cash in” phase and reap the benefits of centralization.

I highly recommend reading some of @atomicpoet​@mastodon.social's write-ups on how this went with Twitter. A couple of references: this thread on Hootsuite's fall, with its closing invitation to build on open protocols, and even more importantly this thread on the importance of third-party developers for platform adoption, and its near-closing reminder about how misplaced it is to put your trust in a commercial enterprise to not screw you over. Anybody setting themselves up to play with ATproto or trusting BS is setting themselves up for a huge delusion.

Don't be misled by its characterization as an “open standard”: ATproto and its ecosystem are effectively fully under corporate control. It's not hard to see how catastrophic that's going to be in the longer term, when considering that even the web —based on much more open standards— failed to resist corporate takeover. I refer you again to @rysiek​@mstdn.social's post and its comparison between the BS/ATproto situation and the monocultures that have and still do dominate —and hold back— the web. It doesn't matter what the spec says. Ultimately use is determined by what the juggernauts in the field choose, and for ATproto that's BS, just like for the web it's GAFAM (and today Google in particular). And yes, this is also a problem with Mastodon in the Fediverse (it's a common complaint, in fact), and even that is not at the scale at which this is a problem with Google on the web, and even that is not at the scale at which this is and will be a problem with BS and ATproto. Forewarned is forearmed.

For users, there is no doubt that the bridge provides a useful service, allowing people on either side to follow and interact with people on the other side, even if at the moment its usage is pretty limited: the bridge has 14.7K followers declared on the BS side, and by my estimation 11.1K on the Fediverse side. (Some would like to change that, pump the numbers up making the bridge opt-out instead of opt-in (a relevant thread). Informed consent? What's that?)

For BS, there is also no doubt that at the moment the bridge is a net win, providing a veneer of decentralization for an essentially centralized service, and giving them a latch on the Fediverse.

For the Fediverse as a whole, the evaluation of the bridge effect is less favorable. By giving BS credibility on the decentralization claim, it damages the public perception of what decentralization actually is, and implicitly supports BS' false claims about the (alleged) superiority of the ATproto approach to decentralization (since it hides that, on the contrary, all perceived benefits in the BS ecosystem come from its centralization instead).

On a larger scale, it will also increase the appeal of BS over the Fediverse for people who would be otherwise more inclined towards the latter for the network effect. In this sense, it provides BS with the rug to be pulled that was already discussed in the previous post concerning Threads' ActivityPub integration (when it will be complete, if not already at this stage of development).

As in that case, while it's theoretically possible for this integration (native in Threads' case, provided by the bridge in the BS case) to work in the other direction as well, helping people move to the Fediverse when (not if, but when) the commercial silos start their enshittification process, it's more likely that such a channel will be plugged as soon as the flow becomes a threat for the silos. In the mean time, it's more likely to favor migration from the Fediverse to the silos, that are generally more appealing in this initial growth phases.

(Of course, for both the Threads and BS bridge case, the argument is grounded in numbers. We won't see any meaningful effect from the bridge presence as long as cross-network federation is low, nor from the Threads integration until it reaches completion and sees wider adoption.)

But while in the Threads case the channel (i.e. the rug to be pulled) is entirely controlled by its genocidal parent company, and there's little that can be done about it Fediverse-side except joining the FediPact against federation with Threads, in the BS case it's almost silly how we're basically weaving that rug ourselves.

But it's worse than that

BS actually has some extra cards to play against the Fediverse, compared to Threads. For example, they can (and do) play the above-mentioned “our approach to decentralization is better” card (even if the only thing that makes it better is … the centralization; but facts don't really matter in the public perception).

Moreover, despite the potential threat of BS pulling the rug (deciding to go its own way) in more mature times, I don't actually think BS will have a particular inclination to cut off external PDSes, except maybe for egregious reasons (not hard to guess one, and it's not “only hosting Nazis”). This isn't so much out of good will as much as a matter of convenience. Again, as also reported by @rysiek​@mstdn.social, several people have noticed that the structure of the network is designed so that BS can offload work to other parties while reaping the benefits of “centralization where it matters” for itself. In this sense, ATproto is the epitome of capitalism (privatize profits, socialize losses), and chokepoint capitalism at that, enshrined at the protocol level (of course, BS is setting itself up as the chokepoint).

And before anyone jumps in, remarking that the protocol doesn't explicitly require the existence of a large centralized server: the reliance on «fairly resource-demanding» Relays and App Views for network-wide data collection, distribution and presentation is the relevant cornerstone highlighted in the protocol documentation itself, and clear to anyone who has given any thought to what it would mean to host an alternative relay other than for niche communities.

(This is essentially a discourse similar to what could be done about the open web and search engines. And not by chance.)

While we're at it, one of the most ridiculous statements in the ATproto design documents is their claim («“Big World” Design» section, that I can't even link directly because apparently whoever wrote the docs doesn't want that particular section to be linked) is the claim that the protocol is «modeled after the open web itself».

Here's the next two sentences:

With the web, individual computers upload content to the network, and then all of that content is then broadcasted back to other computers. Similarly, with the AT Protocol, we’re sending messages to a much smaller number of big aggregators, which then broadcast that data to personal data servers across the network.

This is one of the worst misrepresentation of what the open web is (or more precisely: what it was intended to be) that I've ever seen. It's even worse than the Series of tubes take. The web, especially the open web, was never about “uploading content to the network” to have it then “broadcasted” to other computers. The key defining characteristic of the web was interconnection between independent computers, directly accessible from any other (inter)connected computer. The original design document of the WWW has this to say about publishing documents on the web (emphasis mine):

From the information provider’s point of view, existing information systems may be “published” as part of the web simply by giving access to the data through a small server program. The data itself, and the software and human procedures which manage it, are left entirely in place.

I doubt that the writers of that document made such a glaring mistake by ignorance or distraction. The misrepresentation is intentional, to fake an analogy between the centralizing nature of BS' ATproto and the completely opposite, decentralized spirit and intent of the open web that has been completely destroyed by the growth of GAFAM. Ironically, this misrepresentation aims at presenting BS and its ATproto as a decentralized alternative to Big Tech while presenting the Big Tech chokepoint model as the open web model.

They then go on to present three justifications for the opportunity of what they call the “big-world indexing” model —except that the excuses they present are mostly bullshit (I mean, they're not BS for nothing), and when they're not they still don't require the kind of centralized model that they propose, revealing the actual intent behind such centralization (already discussed in those same 1, 2 and 3). It's also almost offensive how they don't explicitly mention ActivityPub, which is most surely what they're really comparing against.

Let's see them in detail.

Reduce load on PDSs

(“Make it easier to self-host, you can easily run your own server.”)

The open web model that ATproto claims to be inspired by is a “pull” protocol. When someone is interested in a content you publish, they fetch it directly from your server. In an analogy with a social network with a similar model, the load on your server would typically depend on how many followers you have (think RSS). And yes, if you start getting viral and/or have a huge following a 56K home connection might not cut it. There's a reason why slashdotting has become a verb.

The solution to this is to upgrade to a beefier hosting service (especially in terms of bandwidth) if this becomes a routine occurrence, not to move towards centralization for everything and everybody, all in the hands of a single (or a few) corporations.

But it's more likely that the BS document is taking a jab at Mastodon, that is infamously … not particularly efficient to host. However, that's a Mastodon issue more than a protocol issue. (This, by the way, is something you'll hear me say frequently. Some other time I may even write an article about the many ways in which Mastodon is and has been bad for the Fediverse, despite the tremendous contribution it has given to its expansion.)

There are other ActivityPub servers that are considerably more lightweight, especially for single-user (or more in general small-server) use. Pleroma is one such example, whose lightness and easy of installation ended up giving it a bad name because it became the platform of choice for neo-Nazis and related trolls in their harassment campaigns throughout the Fediverse. More recently, GoToSocial has also emerged as a lightweight server, specifically designed for single-user or few-users instances.

Being a “push” protocol, the load on an ActivityPub server depends mostly on how many people are followed by the users on the server For small servers, and especially personal servers, this is generally going to be on the low end. From a poll I've recently run on the Fediverse, the number of accounts followed by each users is measured in the hundreds, or at worst the low thousands. Even if all of them were heavy posters (say, 300 posts per day), this would still lead to 3 to 5 connections per second, easily handled by a home server.

(And that's a gross overestimation, since in the Fediverse it's not uncommon to end up following hundreds if not thousands of accounts simply because many of them go dormant after a while. There are also other cases, such as people who prefer following hashtags rather than accounts, but that's a separate topic.)

And sure, if you follow several media-heavy posters you may want to watch your disk usage if your server caches remote media files (not all servers do), but again, this does not justify designing a different protocol around centralization.

(There's another side to this, which is the infamous Mastodon stampede that can DDoS not only other instances, but in general any website, especially poorly-designed ones, as metadata is fetched for links and images to build cards and previews, turning from “push” to “pull” and going back to the beginning of this subsection. Again, this isn't fixed by a different federation protocol, but by a different way to federate said cards and previews.)

Improve discoverability

That's called a search engine. You don't need a new protocol to make a search engine, unless you want to make life easier for the search engine so that it can lower costs and maximize profits.

The reason why search on the web is going to shit isn't that there isn't a good protocol for it, it's that monetization has taken priority over providing the actual service. Kind of like the enshittification process that has destroyed all centralized social silos, which is exactly the path that BS will take a few years from now, when it'll need to start paying back the VCs to which it has indebted itself.

The reason why search engines and directories specifically designed around the Fediverse are sparse is cultural, not technical: there is nothing preventing a crawler from browsing Fediverse instances and collect information about accounts and posts, and present the collected information in a way that “improves discoverability”. Many Fediverse users though, particularly in the old guard would rather have these services opt-in (informed consent). For the same reason, a lot of users post with more restrictive post privacy than “public”.

This leads to an aperiodic re-emergence of similar situations, in which the techbro du jour sets up a Fediverse scraper that completely disregards this cultural inclination, going for opt-out (if even allowing that at all), on the excuse that “opt-in means few people will join”, and will be put through the grill for not understanding the basics of consent (flash news: if few people are interested in joining your service when it's opt-in, it means your services is not appreciated, making it opt-out is a bad idea, not genius).

This, again, is not solved by the use of a different protocol, especially not one that essentially designed around removing agency from the posters.

Improve “quality of experience”

(“Fewer dropped messages, out of sync metrics”)

This is probably the most bullshit of excuses. A separate discussions deserves to be dedicated to how both are indicative of a design intent that aims at replicating the toxic usage of social media that has become preponderant in the last decade plus, and that we should be breaking free from rather than encourage. The “dropped messages” in particular is FUD, in addition to stoking the well-known FOMO that is instrumental to the commercial social networks' manipulative approach.

And of course, neither of those really need centralization. Talking metrics, for example, even on Mastodon, which is most likely what this point is —again— taking a jab at, metrics are always accurate on the home server, that should be the only one to care about them.

«I don't know how many people liked or boosted this not-mine message.» really isn't something most people outside of the author should have to worry about, and for the infrequent cases when it does matter (research, for example) it can always be fetched “fresh” from the message home server.

Again, not an argument for centralization.

As for the dropped messages … ActivityPub doesn't “drop messages”. I assume that what the document authors are taking a jab at is Mastodon's infamously “conservative” approach to remote profile backfilling, full thread federation, and similar related issues.

Again, this is not a protocol issue, but a Mastodon issue. Even alternative front-ends to Mastodon (such as Phanpy) can give the user access to remote content with relative ease. Other server software, such as Friendica, work largely on the principle of loading remote content on an as-needed basis. There are even Mastodon forks that provide better support for importing more remote content on request.

FWIW, the partial thread federation in particular is indeed one of the most annoying misfeatures in Mastodon, and I'm looking forward to see it fixed —even though I don't think it'll resolve one of the common complaints derived from it (repeated similar replies from different people that don't see each other's replies) —simply because I often saw that same thing happening on Twitter: a lot of people reply to posts without first checking out the thread.

But one thing's for sure, you don't need a different protocol to fix it.

ATproto doesn't actually do anything to solve these issues

What's even worse is that even if these issues were actually protocol issues (which they are not), ATproto does absolutely nothing to fix those. It's not the restructuring of the network in PDS, Relay and App View that “solves” any of those problem, but the fact there is a single huge centralized node (coincidentally, the one provided by BS, the commercial enterprise) that handles them all.

You would get the same effect on the Fediverse if 99.7% of the accounts were on “the Mastodon flagship instance” (mastodon.social), the few other instances federated with it, and Mastodon did a remote-fetch when opening a thread.

The only reason why “the ATmosphere” isn't seeing any of the issues exhibited by the Fediverse isn't magic fairy dust in the protocol, but the fact that BS offers a centralized service covering all 4 fundamental components of ATproto:

1. an identity provider (this article is getting long so I'll postpone the rant about DIDs to some other time);
2. a not-so-Personal ATproto Data Server (PDS) (multiple ones, in fact, to distribute load, but presented as one);
3. an ATproto Relay;
4. an ATproto App View;

(Yes, I'm simplifying a bit here. There are other aspects such as labelling and feed generators. There would be a lot to say about this, but the only thing relevant here is how they contribute to the “let us other work for us for free” aspect.)

Now let's imagine a scenario in which ATproto takes off, BS grows to the point it can switch to the “cash in” phase, and just as it starts down the enshittification path to pay back its debts, some Big Tech competitor sets up its own ATproto service (call it BC): identity provider, hosted PDSes, Relay and App View. And of course, since in the beginning it only has few users, the Relay is set up to slurp in the entire contents of the BS PDS(es).

First question: do you think BS will let them do it? Or would you bet on the BC Relay getting blocked by the BS PDS within a couple of days, if not in a matter of hours?

(Hint: this is exactly what led Google Chat to close XMPP federation: no problem while the federated servers were small personal ones, but as soon as Facebook Messenger cooked up its compatibility layer and started crawling the Google Chat account network, Google isolated their server. That's the rug pull I mentioned already in the last post.)

Heck, even if BS decides to “play fair” and let this hypothetical BC do its thing, how much data would the BC relay need to fetch to start becoming even just moderately … I won't say competitive, but at least useful, maintaining the promise of “fewer dropped message, no out-of-sync metrics” allegedly guaranteed by the protocol?

What if this hypothetical BC experiences sufficient growth on its own PDS(es) (by migration —an as of yet a completely untested alleged feature of ATproto and BS, especially between competing, potentially hostile and reciprocally blocking hosting services— or any other mean) and decides they don't need to share that data with BS? How will ATproto help BS “improve discoverability” when they're cut out from half (or more) of the network because the BC PDS(es) refuse to communicate with the BS Relay?

(Yes, some of these topics are similar or symmetric to the ones discussed elsewhere about alternative relays and PDSes.)

So yeah, none of the claimed benefits of the protocol come from the protocol itself, and the only thing at which the protocol could be better than ActivityPub is the DID and its portability promise that still has to withstand the test of reality.

(Not to mention the centralized PLC DID directory under BS control that they promise will be spun off —while still of course remaining a centralized service: what happens when this directory service decides to block your account because your stuff has too much skin or too many female-presenting nipples? —never for the Nazi stuff, BTW, that's always OK for these services, of course. Sorry, I'll stop, I promised I will rant about this some other time.)

Seriously, if you want a truly decentralized social network where you remain in complete control of your data and connections, you'd be better off with something like Secure Scuttlebutt than with anything in the BS orbit, or with a service like Streams, that supports both ActivityPub and Nomadic Identities —anything, in fact, but that centralized mockery of federation that is BS and its AT protocol.

Stop making excuses for them

I consider the lies and misdirection in the ATproto documentation part of the course —after all, they have to sell their business to buyers, assuming federation will ever be a selling point for corporations.

But what's really surprising to me is the number of non-employees falling for it, the useful idiots they manage to get on board. Every time I see someone who is not a BS employee trying to push the idea that ATproto is decentralized I have to wonder where it's coming from, especially since the design documents themselves repeat in multiple points that the protocol is expressly designed for centralization (pardon, “Big World” Design).

No matter how your extremely cute and/or interesting ATproto-based application can claim to be independent of BS, it becomes essentially useless unless it builds on their relay. Ignoring that is either malicious or stupid, and completely misses the point of both ATproto and BS. Oh sure, you can actually design it to not depend on the BS relay: but then you'll soon discover it to be affected by all the pains of decentralization, and you might as well have built it on any other protocol not controlled by a corporation to sell their centralization services under the pretense of “decentralization”.

Since @jdp23​@gotosocial.thenexus.today woud like you readers of mine to know who I'm insulting in my warning rants against “expansive” definitions of the Fediverse or even of the concept of decentralization (because they are well aware that by no current definition of the term BS can be considered decentralized, so they need to resort to etymological jungle-gymming to make the entity fit the mold, not unlike the aforementioned libertoloids1 with the concept of freedom or voluntarity to justify the exploitation of people in need3), I feel the need to quote George Santayana here, which has some irony to it4. The following excerpts are from the first volume (Reason in common sense) of his Life of Reason, which you can read in full on Project Gutenberg.

First and foremost, his most famous and frequently paraphrased quote:

Those who cannot remember the past are condemned to repeat it.

There's more to it though, and I feel that the next paragraph makes a better point of it:

Not all readaptation […] is progress, for ideal identity must not be lost. The Latin language did not progress when it passed into Italian. It died. […] when the foundation itself shifts […] progress is not real. […] without this stability at the core no common standard exists and all comparison of value with value must be external and arbitrary. Retentiveness, we must repeat, is the condition of progress.

I was also reminded that Evan Prodromou, considered “the father of the Fediverse”, includes BS in the Fediverse, despite the fact that BS doesn't implement ActivityPub at all —something that is otherwise apparently enough to disqualify diaspora*— and the only reason it can communicate with the rest of the Fediverse is because of the bridge —which is a bit like saying that Tumblr is part of the Fediverse because Friendica can bridge to it (like it can to diaspora*).

But then again, Friendica, like most other Macgirvin projects, was always the odd-one out in the Fediverse, despite sporting much-needed features (first and foremost that nomadic identity that ATproto claims as (untested) advantage over ActivityPub), they never managed to gain much traction, and the feature themselves have remained ignored at large. And as also @jdp23 points out, this was one of the excuses used by the ATproto developers to recuse ActivityPub5 —which of course, as excuses go, it's a pretty piss-poor one, since once again it doesn't justify building an entirely new, untested, protocol from scratch rather than helping the convergence of the existing ones (which is technically possible, if there's a will to do it).

(Of course, we know why they didn't go that way, and in this case it's probably not even CADT, but a matter of intellectual property, budget justification, and all the familiar reasons that fuel the worst in corporate-driven “open source” development.)

Anyway, I'm not entirely sure what Prodromou's endorsement of BS is supposed to change. Am I supposed to be impressed by an appeal to autority? For an apt comparison, Tim Berners-Lee (the Father of the Web) —for whom I have way more esteem than for Evan— completely burned all credibility when he endorsed the W3C “standardization” of Digital Restrictions Management for the free and open web they were supposed to promote.

This kind of regulatory capture is exactly what's on the horizon for the Fediverse, as I've been saying for a while now (unsurprisingly, Prodromou has my Fediverse account blocked since the time I started actively posting to warn about said threat to the Fediverse). Why exactly should I waste my time mincing my words? I know where his enthusiasm is coming from, and still think he should do better than seek for recognition via the numbers game.

The bait

With this all being said, as I mentioned before, I'm convinced that BS will be generally welcoming not only of third-party “applications” (especially in the beginning, where they have to build momentum), but especially (and probably for longer) of external PDSes.

One reason that I haven't mentioned yet (OK, that I teased at the beginning) is that it gives them more material for when they'll start looking into the SALAMI grift for cash. Thanks to the unavoidably liberal license that users grant them on the content they share on the network, they'll get free hand on using it as source material to train their models. Even better, when they start on this path they'll be able to do that without any public announcement to ask for permission, since they can formally delegate that to a submarine spin-off or pretend-independent initiative that just happens to be build an “App View” (read: a leech on BS' relay's “firehose”) that is a SALAMI training setup, and by the time it'll go public it'll be to late to do anything about it.

Although this isn't a direct threat to the Fediverse by itself (in fact, it might even encourage more people to finally appreciate the value of an independent internet, and give more weight to the benefits that come from the pains of federation), it poses an interesting question about the viability of the bridge, given the generally higher sensibility users on the Fediverse have around these topics. Keep the bridge open, to encourage people to move to the Fediverse (the SALAMI grift is one of those enshittification steps that can really help drive out the artists, for example), but potentially providing data for said grift, or start cutting it off to protect against the grift, but making it harder for potential fugitives to migrate?

(On the upside, the approach adopted by the Fediverse bridge to rely on follow relationships as sign of consent gives the network members some control on what and when data flows through the channel. Of course there's people who'd like to change that, but hopefully if this does change it'll only be on BS side.)

Threads as the “backup plan”?

All things considered, at least until the famous account migration finally gets “battle-tested”, there is only one sense in which ATproto is fundamentally different from ActivityPub: it's more corporate-oriented. While this is mostly something that matters only “server-side”, it will most likely reflect in the culture that will develop in its space, which will be a clear positive for people with a specific mindset (to wit, the kind of people that find the mutual aid posts in the Fediverse “offensive”): there's no doubt they'll see the protocol as “better”: not because it does decentralization better —it doesn't; in fact, it does it much worse— but because they'll be able to look at it as a platform they can exploit to their benefit —at least until they get bitten in the back by BS' own monetization about-turn.

(… taking their users with them. See also Cory Doctorow's take on why he's avoiding BS, and for balance Molly White on the fact that she's going to establish a presence there anyway. Fun fanct: since he's published this article, I've been seeing more posts with negative takes on Doctorow or some of the terms he has popularized, like “enshittification”. Which may be just a coincidence, or maybe he just touched a lot of exposed nerves with this one.)

The measure in which this is a threat to the Fediverse is proportional to the amount of people BS manages to convince of their lie about the “decentralized” nature of the ATmosphere (with them at the center, of course). As I mentioned, the reason this is a problem of the Fediverse is that it projects the false impression that every single issue people come across in the Fediverse are a specific limitation of that particular network, rather than, as they are for the most part, a general issue with decentralization. And while it is true that the Fediverse has issues of its own (among which, sorry to say this, the predominance of Mastodon, but I recommend reading some of the stuff written by @trwnh​@mastodon.social to get an idea about all of the corners of the ActivityPub specification that needs some solid work, or the discussion up on the SocialHub to see how discussions about them are handled) the ones most commonly encountered are only solved in BS because of the centralization around their corporate nodes, and not by virtue of their choice of protocol.

That being said, I would greatly enjoy to see a couple of Big Tech corps pick it up just to enjoy the fantastic mess that will emerge from them tripping over each other in an effort to steal each other's lemons fodder esteemed users to maximize the profit from the leeched data resale. But I suspect this won't happen any time soon, so we'll be left to suffer through a constant stream of lies about the untestable claims of ATproto's superiority.

So what does this have to do with Threads?

At least in theory, the attention that Threads has been giving to ActivityPub through their (partial, opt-in) support for federation could just as well extend to ATproto. (After all, who better than them is in a position to do that, given their extensive computational resources?) I doubt this will happen any time soon though —if ever at all— and the ATproto being still in infancy and largely untested (especially in terms of federation), not fully specified, and controlled by a direct competitor are only the first reasons the come to mind.

(FWIW, ActivityPub itself is substantially underspecified, but it at least enjoys several years of practical experimentation that has defined de facto solutions for some of its less clear points.)

How much is Threads actually interested in federation, though? Regardless of the good intentions of the engineers working on it, does the genocidal parent company actually think it's a goal worth pursuing, or are they just looking into the bare minimum to work around any constraints that would otherwise come from the European Digital Markets Act?

(By the way, this is also another reason why they are unlikely to federate with BS via ATproto: it's more strategic for them to let BS grow until it can be declared a “gatekeeper” (for its own network) under the DMA, and then look into opportunities to exploit there.)

There has been a missed opportunity with the more recent exodus from Xitter. BS has clearly benefited from this (that bump in the stats is incontrovertible), and while a smaller bump was also observed in the Fediverse (visible on both the FediDB and Fediverse Observer stats), there's little doubt that it (the bump, but also the Fediverse in general, and Mastodon in it) have seen much more fanfare compared to the 2022. In all this, not only it's not clear how many users Threads itself has gained from this migration (it's surprisingly hard to get any decent historical stats of Threads usage —or at least, I haven't been able to find any), but it's also quite clear that the company hasn't been exactly pushing hard on its support for federation with Mastodon (let alone the Fediverse in general) to appeal to fugitives.

There are obviously several possible reasons for this, other than the obvious “Threads is only paying lip service to the Fediverse, doesn't really see it as a selling point now that the novelty has worn off, and wants people to end up on Threads, not join the Fediverse in general, so of course they won't do anything that gives the Fediverse credibility”.

For example, there's the fact that federation as a concept in general is still widely unknown and misunderstood (which is how BS can sell their claim to be decentralized), and this widespread ignorance contributes to it not being a selling point (aside from the FUD about how “it's complicated” that BS rides to claim they do it better).

And while it's also quite possible that the reason for the silence is that there simply haven't been any significant changes to the federation support since the end of August, it's telling that they don't seem to be playing this card at all, even in face of the threat of migration off to BS due to moderation issues (something to the tune of “If you don't want to join Threads, at least stick to the Fediverse so you can still follow the people you want to follow that are on Threads”)

Telling, but not surprising, also because those moderation issues keep hitting denouncing Trump's Nazi sympathies “by mistake”, while hate speech from the other side gets promoted “by mistake” —this of course has absolutely nothing to do with the hiring at Meta of one of the authors of the neo-Nazi Project 2025 driving current Republican politics— whereas the rest of the Fediverse is strongly left-leaning.

Talk about Threads federation being a taint on the Fediverse —and then people are surprised by the FediPact. But don't worry. In the mean time, Meta has joined the Social Web Foundation, putting itself in a more solid position to manipulate the future development of the protocols and platforms on which the Fediverse is built, steering them towards more corporate-friendly pastures and away for surveillance-capitalism-resistant initiatives.

(Of course, this is still only the preparatory stage: we're barely at the beginning of the Embrace step in EEE. What really matters it what will come next.)