Mandatory disclaimer: I’m writing this piece as a personal opinion. It only reflects my views and not those of my employer (Meta — though the area I work is not directly related to the topics at hand) who has not had any influence on my writing this piece. I also neither have nor have had any stake in Twitter or other social media companies beside my employer, to be clear.
In my previous post on social media, I have pointed out that I didn’t think that Mastodon would take on as a valuable Twitter alternative if the latter did collapse. Given that we’re definitely in the “collapse” process right now, we’re seeing a significant influx of users into Mastodon, which I, admittedly, I’m enjoying. On the other hand, I’m still skeptical about the its long-term viability.
The first problem, in my opinion, is that everyone is bringing a different view of what they want out of this migration, both between the early adopters and the newcomers. Some of them are considering Mastodon a straight-up Twitter replacement, others insist that they should “learn their manners” and that Mastodon is most definitely not a replacement for Twitter, with a general handwave about “the Fediverse.”
This, I foresee, is likely going to be the source of misunderstandings, fights, and schism. It’s somewhat reminding me of overly simplified Buddhist views of Hīnayāna and Mahāyāna, with some people expecting to be able to run a global network “with kindness built-in” while others are expecting a throwback to the early Internet with communities keeping by themselves rather than being brought onto a global plaza.
Personally, and once again I repeat personally, I would much prefer keeping the global square, mostly because I don’t fit in communities due to my tendency to criticize my own view of the world. This is also one of the reasons why I can’t think of a better option for myself than staying on the “default” Mastodon instance: I would likely be in breach of a norm, written or unwritten, of not disagreeing within the community, I definitely did that in the past with FLOSS.
And while there is portability built into the whole federation idea, there also is an issue of building said federation. Linking across instances does not happen by default – just like with XMPP back in the days of Jabber! – and while it should be straightforward in most cases, admins can reject a federation request. And while some cases are clear cut, there are already cases surfacing of instances being picky with their federation requests with reasons whimsical at best (hopefully, many will be straighten out now that a possible critical mass is being built.) Again, I see enough people liking this isolationist approach, I personally don’t.
The second problem is closer to my professional views and that’s where things are harder for me to express safely. The short version is that I’m afraid there’s little chance of the current architecture to scale up to the point that it needs to, to be able to sustain the amount of traffic that Twitter has had up to now.
People already are discovering the costs behind running a single-user Mastodon instance when following and being followed by thousands of accounts. Influencers and personalities who expected thousands or millions of followers will have to pay for hosting and for a sysadmin team to keep them operating, particularly if their presence on Mastodon is revenue-critical. Which makes a good point for organizations such as BBC, NatGeo, FT, and so on to run their own instance as part of their infrastructure — until anti-establishment instances will decide to block them en-mass. because they can.
The requirement of back-and-forth between followers and posters means that the distributed system that Mastodon instances are part of (the so-called Fediverse) does neither get the economy of scale of sharing the load across different datacenters throughout the world, nor it distributes the load on different instances — adding more instances only increases the traffic on the network.
And this has me segue into a third problem: funding. Yes, most instances will have a Patreon to cover the infrastructure costs. But will these ever get to cover the needs of a platform to be as successful as Twitter? Just consider that, for a split follow-the-sun 24/7 rotation, Google needs 12 engineers – any less than that, and the oncall rotation is not established or is merged into a different one, as it means everyone is oncall for around a week every month and a half – even putting them at a discount given the current crisis and the layoffs, let’s call them $125k/yr. That means for a Big Tech company, running the minimum service costs $1.5M/yr at least — and sure, most of the people on these rotations are also working on the development rather than being dedicated ops, but assuming that an instance admin also works on building features or moderating the content, it should be roughly equivalent.
And yes I can already hear all of the proclaim of doing it for the community, volunteering on spare time, and so on. I get it, it can work for a while, it can work for some instances, they can even become cooperatives (some of them already are!), but I do not believe that, in the long term, this is sustainable. It’s a similar risk of being a successful open source developer: once the market for SRE-type roles picks up again, those who have worked for long enough on ensuring these instances keeps working reliably will have the street creds to be hired a lot faster. Which is why, even though at this time factoring in the admins as full-time engineer salaries is pointless, it should be considered on the longer term.
The next problem in my mind after that has also to do with scaling, but from a different point of view: while regulators are unlikely to put a small network of federated instances under scrutiny, if a critical mass of people will switch to the Fediverse, they will get noticed. Much as some of the folks behind this will likely appeal to the declaration of independence of cyberspace, even cryptocurrencies encountered regulation, eventually, and the same will happen to this type of federated systems. What this amounts to will change depending on where the instances are hosted, where the admins live, which passports they have. From basic PII handling (GDPR compliance) to libel laws, from DMCA to CSAM, I seriously doubt that volunteer content moderators will be able to arrange for them without the help of legal support — and given liability, I doubt there’ll be a lot of lawyers queuing up to volunteer for it. And honestly, I would never have the mental resilience to become a moderator for a service that allows users to post whatever media content they want — and even much smaller services than this had their share of interactions with law enforcement because of that.
Now, a lot of these problems are the kind of problems you can solve by throwing money at them. That’s what happens in Big Tech in my experience. But to get that kind of money you usually only get two options: you get VCs to throw it at you (unlikely, in the current financial situation), or you make that kind of money, mostly through ads. I wrote already years ago about my opinion on internet ads and even already discussed how I think ads are better than subscriptions given the pay-to-play nature of paying for services that should provide a public service. And not because I happen to work for a company that serves those ads, just so you know — my opinions on the matter are publicly recorded in this very blogs over the years and I believe I had the same consistent view on this since I started blogging.
But ads, in Mastodon? Not going to happen — they could happen on some specific instance, but that’s extremely unlikely that users would stick to such an instance, and for advertisers to cater to such a small pool of users. And generally advertisers ask of their media agencies some data to prove that whatever they are doing is even effective — with such a privacy-first approach that ActivityPub (the protocol) and Mastodon (the software stack) have taken, this is unlikely to happen.
To take the simile that many are making, “Mastodon is like email” — for most cases, hosting your own email is a bad deal already, but this brings another problem: most people are unlikely to care about paying for good email hosting. I love my Fastmail and pay for it, but most people jumped on Gmail because they gave a reliable, solid email service for free — before it the vast majority of the people I knew just used whatever their provider gave them, which was usually only reachable from within the provider’s own network anyway! And even businesses that depend on email (and other Google-provided services) for their income tend to not sign up for the paid editions — just think how many restaurants have an @gmail.com address! And that’s without going down the weeds of spam handling, security, etc.
Am I suggesting, or even hoping, that Mastodon is going to fail? No, at least definitely not hoping. I do not see how the whole Fediverse will survive long term though — it has built-in scaling issues in both technology and relationships. This does not mean dismissing it altogether, it means pointing out the holes so that they can be addressed — even where it is difficult.
I know that the “pioneer spirits” out there would rather see the big commercial institutions fail, but I personally expect that what will end up getting traction is a handful of “high polish” instances recreating Twitter, possibly with opt-in basic analytics (with significant messaging to recommend people to opt-in), and the ability to pay for zero-ads experience, and limited federation, plus a number of media company instances (likely white-label, maybe directly from the advertising agencies) carrying different level of one-way posting — providing support by DM, which is very common on Twitter nowadays, would not work when multiple third-parties are involved!
This does not mean I believe this is the correct implementation, but just the most survivable in the world we are in right now. I will be glad to be proved wrong.