Official install guide is underlining that all the Cloudflare's offerings are still a disjointed bunch. You need to enable billing in 3 separate products in 3 different UIs (while only Images requires an actual paid plan, Access requires a credit card to onboard and Workers Unbound is probably a good idea to enable right away).
Why so much management/onboarding complexity if you really want to compete with AWS as a general-purpose cloud?
I honestly believe that large instances (thousands of users) are terrible for the fediverse (due to moderation, security, cost, and fracturing due to domain blocks) and that more people / small groups should be running their own instances.
Was wondering what HN thinks about this solution, and if you run your own instance of an activity pub / fediverse server that isn't mastodon which are you using? I've been trying to find something that is easy and clean for a single user instance that could run on a raspberry pi or other tiny low powered setup.
> if you run your own instance of an activity pub / fediverse server that isn't mastodon which are you using?
I've got my main instance as Akkoma[1] (was Pleroma, both in Elixir), a secondary as GotoSocial[2] (Go), and a couple of Honks[3] (Go) for various other things.
This seems really great to me. Even with Digital Ocean's pre-baked Mastodon deployment, there's still a good amount of operational overhead and cost. People lament centralization, but there's plenty of choices for people who want various levels of independence.
The Fediverse is like email. There's Gmail and other completely hosted providers. I'm sure there are services that are out of the box but deployed to cloud infra. And you can self-host if you really want.
I hope a lot of the big players enter the space. That would help prevent embrace-and-extinguish.
It's interesting how much it costs.
Also, it's interesting is it possible to migrate to fully-self-hosted Mastodoon setup even without data (it's very difficult in Matrix - as far as I understood, you can't migrate Synapse-based instance to non-Synapse and you can't just start clean on same url)
I think a company like Twitter could have had a real competitive advantage if they were to have repackaged its core experience as a "bring your own domain name" hosted implementation of ActivityPub in the likes of G Suite or M365. I thought for a minute that's what Bluesky was, but I am losing all hopes they'll have sense to do that.
This is not a Cloudflare-run activitypub server. This is software that you can deploy on Cloudflare (or, potentially, on top of open-source workerd) to run your own server. Creating more implementations of ActivityPub and more places you can deploy it is what the fediverse is all about, don't you think?
(Disclosure: I am the tech lead for Cloudflare Workers.)
> Is there working stack you can deploy to your own VM and deploy Wildebeest to?
I don't think anyone has tried it yet. There are probably some missing pieces that need to be filled in. In principle, it should be possible to fill in those pieces without changing any of the Wildebeest code itself.
As a simple example, Workers KV provides a simple key/value storage API, basically just the methods get(key) and put(key, value). In workerd, you can use this API, but you have to configure it to point at your own back-end that does the actual storage. It translates into a basic HTTP API (mostly just GET/PUT). You could pretty easily write a Worker (running as another nanoservice on top of the same workerd instance) that uses local disk storage to back this API, or sends the data to some other back-end of your choice.
Eventually we should have ready-made shims like this to handle a lot of situations, but it's still somewhat early days of workerd and at present some of this is still missing.
Official install guide is underlining that all the Cloudflare's offerings are still a disjointed bunch. You need to enable billing in 3 separate products in 3 different UIs (while only Images requires an actual paid plan, Access requires a credit card to onboard and Workers Unbound is probably a good idea to enable right away).
Why so much management/onboarding complexity if you really want to compete with AWS as a general-purpose cloud?
(edit: spelling)