Twitter Bluesky, Protocols versus Platforms and the Fediverse
Twitter is funding a small independent team of up to five open source architects, engineers, and designers to develop an open and decentralized standard for social media. The goal is for Twitter to ultimately be a client of this standard. 🧵
- jack 🌍🌏🌎 (@jack) December 11, 2019
Under the name @bluesky, the crux of the idea is to define an open protocol that many vendors could implement and which Twitter would also support.
Doing this would certainly be a major positive: breaking the dependence on one company and infrastructure for public discourse has huge advantages. By opening up preemptively, Twitter could likely still remain the major form of access for most users even if others implemented services. Hopefully, the team will be well resourced and make plenty of progress.
One of the unfortunate things about the announcement though was that it failed to acknowledge that there are already a slew of open protocol based social networks and other services out there that are taking exactly this approach. (Adi Robertson’s post at the Verge does a great job of summarizing their reaction at the announcement).
It’s Tiny, but the Fediverse is Growing
Hopefully, Twitter’s bluesky team will look at the protocols and services already out there and multiply those efforts, but even if they don’t the good news is their announcement has raised the profile of that existing work.
The W3C Activity Pub Protocol isn’t that widely known, but it’s been gaining steam. ActitityPub forms the core technology for a wide range of social network services where the same functionality can be offered by many independently operated servers: creating a disturbed mesh of implementations.
The largest network to date is Mastodon, which provides a short updates service very similar to Twitter (you can find me on Mastodon at firstname.lastname@example.org). Mastodon has more than 2Million users across 60,000 or so independent servers. While this doesn’t sound like much compared to twitter’s 126Million users, especially given how important network effects are in creating value, it’s already an impressive achievement.
Not only that, there are a whole host of fledgling social network services using ActivityPub:
This emerging Fediverse of decentralized services is still in its infancy but the services are genuinely functional. Those wishing to escape centralized platforms are finally getting some genuine functionality there.
Protocols versus Platforms
The key point behind these services is that they are based on open protocols that anyone can implement. Once implemented, new nodes can join the global network and enable communication with users on other nodes. Each node can survive independently of the others and (in theory) set its own rules for usage, access etc. This contrasts with the platform model where one company (Twitter, Facebook, etc.) controls all of the servers that make up a particular network.
The upside of the platform approach is that one provider can provide a reliable consistent experience, implement a functioning business model at scale and will often have a critical mass of users that makes the network valuable.
The downside of the platform approach is that this same provider can make arbitrary choices about what content is permitted, who can use the service, and also has effectively omnipotent control of the economics on the platform. (**)
In the long run, protocols are a greatly superior approach to building planetary-scale systems. It is hard to imagine how the early Internet, the World Wide Web or email could have succeeded if it had tried to follow the early AOL-like walled garden platform model. The opportunity for many players to emerge and compete but interoperate came from open protocols.
A great piece by Mike Masnick digs deeper into how protocols vs platforms also play out for free speech (there are pros and cons, but ultimately protocols offer a more robust set of options over a collection of platforms.)
One of the negatives of the reaction to Twitter’s Bluesky announcement was a significant amount of backlash in “we’re already doing it” from some of the Fediverse advocates. It is natural to see some “it’s about time” reactions and some wariness of an organization as large as Twitter potentially not having the same objectives. Despite this, it would genuinely be better if everybody started assuming good intent at the beginning.
There is time to resist if things go well, but starting with collaboration in mind perhaps has the opportunity of accelerating Twitter joining up with something like ActivityPub. Pushing back immediately makes it less likely they will want to engage.
While the Fediverse will no doubt grow on its own over the years, I’d argue it would be way better if the larger platforms joined the protocol efforts and connected up than staying separate. This will depend on everybody’s behavior.
There are also two very big problems to solve…
The two big problems in decentralized social
Decentralizing social networks removes a key dependency on a key player, but it doesn’t suddenly make the internet a “friendly place”. Nor is it obvious how to support all these decentralized nodes.
- Problem 1 — business models: the wild frontier nature of the decentralized efforts has often led to a blanket rejection of business models such as advertising as “evil”. This, in turn, has created a reliance on donations. It’s possible this will work for some services, but it’s unlikely to be a long term solution for everyone. If these services are to be robust for non-technical users, they will need teams of people to maintain them and this takes stable forms of income. In my view, a protocol based world should allow more freedom to experiment with business models not less. In other words, one could easily imagine network servers that charge for access and others that support free services with advertising. This allows users to choose. Without this freedom, it seems likely that the availability of these services to many will stay out of reach.
- Problem 2 — reputation and regulation: there has been an enormous amount of criticism of Twitter and Facebook over how they have dealt with hate speech and fake news. Some arguing that they restrict speech. Others arguing that they do not police it enough. This is an extremely tough problem to solve. For a while, decentralized networks like Mastodon were presented as “Twitter without Nazis” but it has quickly become obvious that it is, in fact, even harder to control the use of a platform and open source technology for hate speech than it is a single provider platform. The whole ethos of the protocol approach is that anyone can run a set of nodes and be free of controls. So when the hard-right social network Gab migrated to Mastodon this caused the Mastodon communities to have to think hard about their policies. Ultimately the solution taken is the only reasonable one given the network structure: to allow individual nodes to set policies about which networks they accept messages from and then set policies on types of speech. This means they cannot stop Gab from operating, but they don’t have to accept messages from the network. This solves the problem to some extent (and maybe it’s the best that can ever be done) but it really underscores the second key decentralized network problem: it is decentralized reputation which is really hard, not decentralized communication.
Both of these challenges for decentralized social networks are really significant.
The first seems likely to be something that will change over time with experimentation. The doomed App.net paid advertising-free social network failed to take off since too few people were willing to pay. This, in turn, meant that it failed to grow a significant user base. With an open protocol though, one could perfectly imagine a diversity of Mastodon servers, some of which are advertising-funded, some of which are donation funded and some of which charge a fee to be advertising free. All three sub-communities get the benefit of the whole mastodon network and hence justify the value of each of their models.
The second problem is much harder. The danger with simply restricting traffic from certain networks is that this has the potential to break up the public space of discourse. Too much fragmentation and there is no longer one global view of what is happening. This phenomenon is already very real in the United States with different media blocks emphasizing highly partisan messages. A highly fragmented social network space could make this even worse.
The need to rebuild reputation and trust in opinions amongst public figures and individuals is great and it will take some creative solutions. In this regard, while the big platform players are taking a lot of criticism at the moment, they are at least providing one unified public square view.
Mastodon’s Server Covenant does seem to be a good step towards trying to provide a shared view of what is acceptable across servers. Hopefully approaches like this can help solve the problem.
Will the Fediverse “Kill the Platforms”?
If we protocols and decentralized services grow will the large platform players be displaced and cease to be relevant? Or in more bombastic terms “will Mastodon kill Twitter”?
There’s no doubt that the availability of alternatives that are all interoperable will help keep the big players more accountable. Still, it seems unlikely that Facebook or Twitter will be significantly threatened by any of these protocols in the near or medium term.
This isn’t just because these platforms still need to build critical mass but because user preferences are not just about interoperability: users still care about usability, service availability, who is on a platform and so forth. Email is an instructive example here. The SMTP protocol has been fundamental to the Internet experience since it was introduced in 1981 and it still powers email today. Anybody can run an email server and send/receive email. Many tens of hundreds of thousands of email service providers exist and many people ran their own software email servers. Yet today, while all these options still exist, the vast majority of users now use one of just a few hosted email services.
Email client usage is just a skewed towards major providers with 27% for Gmail, 27% for Apple iPhone, 9.1% for Microsoft Outlook. (***)
What this indicates is that even where protocols exist, large players who provide a consistent experience can still succeed. In the late 90’s the dominant consumer email services were AOL, hotmail and Yahoo mail. Gmail executed a spectacular market takeover with its Gmail Service.
It is possible that decentralized social networks will have a big negative impact on Facebook and Twitter. However, it seems more likely that if Twitter and Facebook eventually interoperate with these growing networks, they will still remain the entry points of choice for most mainstream users.
This doesn’t mean decentralization is a losing/wasted proposition though: the availability of choice will certainly keep large players more accountable. It will also genuinely mean a greater diversity of services and a more robust system overall. Regulations such as the EU’s GDPR rules and now California’s CCPA are also forcing platform players to open up private data which increases the chances people an export it for use in other networks so some convergence seems inevitable.
In the meantime, whatever becomes of Twitter Bluesky, we should thank it for bringing the decentralized, protocol-based social network scene some more visibility. Lots of exciting progress has been made in the past few years!
(*) As a side note, the fact that we have easy to remember domains for all these services seems to be about the only positive effect of new IANA TLDs.
(**) Note that this does not mean they will necessarily retain every benefit for themselves. Unless they enable others to make money on the platform, they are unlikely to succeed. However, they have control and can change the rules with no notice to others.
(***) Adding in Apple iPad and Apple Mail would add another 8% and 7% for Apple respectively. Yahoo comes in at 6.1%.
Originally published at https://area67.org on January 26, 2020.