Cloudflare’s Wildebeest Not Looking to Upset Mastodon Users

Last month, Cloudflare launched a new service called Wildebeest, allowing Mastodon users to set up and run their own instances on its infrastructure. Wildebeest uses a laundry list of Cloudflare services, including Supercloud, Workers, Pages, Durable Objects, and Queues. It’s very much about showcasing Cloudflare’s own products, but (as with Fastly) it is also encouraging to see cloud platform companies getting involved in the nascent fediverse.
To find out more about Wildebeest and how Cloudflare sees Mastodon and the fediverse evolving, I spoke to the company’s CTO, John Graham-Cumming.
In the email pitch I received from Cloudflare, it was noted that Graham-Cumming had been “dogfooding” Wildebeest — and sure enough, he set up his own instance last month. He is much more active on Twitter, though; and with 28,000 followers on the bird site, you can’t really blame him. But I was curious to know what he thought of decentralized apps like Mastodon, and whether he thinks it will eventually usurp the centralized Twitter?
“My suspicion is both end up living side by side, and I think they do appeal to different user groups,” he said, citing the much easier sign-up process of Twitter as an example of its “default public, default everything” nature. He compared Mastodon to “the old world of forums,” where there are different groups for different types of users.
Sidestepping Controversy: Search and Content Moderation
The forum-like nature of Mastodon perhaps explains why search is such a hot button issue within its community. As I’ve noted before, you are encouraged in Mastodon to use and search for hashtags. However, if you enter a keyword or other search query — say “Wildebeest” — you will get limited results. It depends on the instance, but it’s fairly common to only get results from your own posts or from the conversations you’ve had (in other instances, you don’t even get that). This is a marked difference from Twitter, where you can search for keywords and get results from across the entire network.
Mastodon’s creators decided to limit search deliberately, in order to avoid trolls using the functionality to spread their hate. But it’s also very frustrating for journalists and others who like to actively search a network, for research and discovery. So I asked Graham-Cumming whether there is any functionality in Wildebeest to make full search possible?
“It’s something we talked about doing, but we haven’t done it so far,” he replied. “I mean, maybe at some point we will think about adding it. It’s something that actually the fediverse itself needs to solve, right, because things are federated. They’re not centrally maintained, like in Twitter.”
What he’s getting at, I think, is that any instance could choose to block your instance — de-federate from you — if you try to add wider search functionality. Others have attempted to index Mastodon content before, as this Hacker News thread illustrates, but they are usually quickly cut off from the wider network.
When I mentioned on my Mastodon account that I would be interviewing Cloudflare’s CTO, several people wanted me to ask him about Cloudflare’s approach to moderation on the fediverse. This is a hot-button topic in web hosting in general, but of course it’s particularly important when hosting a social media site.
“I mean, I think it depends on who has responsibility for the content,” he replied. “And we have a different policy depending on whether we are hosting it, or whether it’s passing through our reverse proxy. So, for example, things like hosted on Cloudflare Images, which Wildebeest uses — there is actually a policy around that. And our trust and safety group look at the content if there are reports about the content that’s on there, which is […] handled slightly differently [than] in the reverse proxy case.”
He added that because Mastodon is open source software, Cloudflare is “not ourselves going to be getting into the moderation of individual instances of Mastodon, it’d be more around our abuse policy and where there’s content which we are hosting.”
Vanilla Mastodon
Wildebeest seems to be a fairly vanilla implementation of the Mastodon software — as noted above with the search example, it hasn’t ruffled any feathers yet (other than an early bug that allowed direct messages to be viewed in public, which has since been fixed). Graham-Cumming confirmed that the inspiration for Wildebeest was straightforward.
“We thought, well, this is open source, and in particular as the API’s for Mastodon are public, we could build something that was compatible and make it something that was easy to deploy on Cloudflare, as we have a developer platform.”
The launch blog post stated that “at the core, Wildebeest is a Cloudflare Pages project running its code using Pages Functions.” These functions, in turn, give developers “full access to the Workers ecosystem.”
“Cloudflare Workers is a developer platform where developers write in JavaScript or TypeScript or a language that compiles to Wasm,” Graham-Cumming explained. “We deploy it across Cloudflare’s network, which is around 280 cities worldwide. And we take care of scaling it automatically, as demand ebbs and flows. And so it’s a way for people to build very, very low latency applications, because we’re very close to where the end users are — but equally, scale them simply.”
Despite the sophisticated technology that Wildebeest was built on, it doesn’t offer any extra functionality for helping people migrate their data from one instance to another.
“The reality is the data is stored in our database, D1, which is SQLite underneath,” Graham-Cumming said. “So I think it would be possible, but we don’t have a migration system set up.”
He then asked why a Mastodon user might want to migrate their data from one instance to another, and I explained that it’s fairly common for people to jump to a new instance — for example, to go from the relatively mainstream mastodon.social to a more niche instance (say, hachyderm.io, which has become popular in tech circles). But when you do that, your old data stays on the previous instance.
The Edge
Finally, I couldn’t resist asking about how Cloudflare defines “the edge” in cloud computing nowadays, since the term is heavily used by its competitors for various use cases — everything from Jamstack architecture to modern JavaScript techniques that use a mixture of server-side and client-side code.
“So we’ve actually stopped using that term, pretty much completely,” he replied. “Because, for exactly the reason you mentioned, which is it’s a bit hard to know what it means. It’s almost like ‘the edge’ is just one component of something you’d really want to have, which is, you’d like to be able to just deploy your code and have it run everywhere.”
Cloudflare sees itself as a development platform, he added, so he thinks talking about “the edge” is limiting.
“We tend to think of our thing as just this big global network, a single cloud, a single availability zone across the world, and let us take care of where the data and the code runs.”
Indeed, that is what Wildebeest is for Mastodon — a hosted service to run that particular open source software. I wish it was a bit more ambitious in what it could enable on the fediverse, but regardless it’s a handy tool for anyone who wants to set up their own instance.