Decide on a ActivityPub architecture

[saerdnaer]

In order to make videos on media.ccc.de discoverable within the Fediverse, we have to decide how to implement the ActivityPub protocol:

Every conference/lecture series on media.ccc.de should be it's own actor, which users in the Fediverse can follow and respond to. We have do decide if the delivery of the individual ActivityPub into inboxes of subscribers should be part of the voctoweb Rails App itself, or if we can delegate the delivery to an existing ActivityPub server implementation. This issue is a collection of possible resources.

Libraries:

• https://fedify.dev A TypeScript library for building federated server apps powered by ActivityPub and other standards (++)
• https://gitlab.com/experimentslabs/federails – an engine that brings ActivityPub to Ruby on Rails applications
• …

Servers:

• https://codeberg.org/superseriousbusiness/gotosocial/ https://gotosocial.org an ActivityPub social network server, written in Golang
• https://codeberg.org/grunfink/snac2 A simple, minimalistic ActivityPub instance written in portable C
• Mastodon?

Other implementations

• repco-activitypub

[jfietkau]

Stating the obvious, Peertube and potentially Owncast are relevant inspiration for video via ActivityPub.

Delightful Lists has an extensive categorized list of ActivityPub projects of all sorts, as well as one for development resources.

I have not personally used any ActivityPub frameworks for Rails or even Ruby. Fedify is wonderful for working in TypeScript and maybe for inspiration if you're going to need to build your own abstraction layer. Using a self-sufficient server as a sort of local proxy with an API can work well too, FediMeteo does that with snac. Mastodon is not great for that purpose, since it does a lot that you can't switch off besides just ActivityPub delivery. Whichever server you choose, it should probably be one that you can get under the hood of if you need to (and you will probably need to). Which means don't make it snac if you'd struggle to work with C code.

ActivityPub is an interesting mix of straightforward and complicated. In my experience, people tend to oversell the difficulty of getting something basic off the ground, especially if you're not keen on building a full-blown social network. But there are also a lot of details to sort through if you want to build something that can do everything people will expect from you. For example, the FEP repository keeps a list of optional feature specs of varying degrees of popularity. Some of these are very esoteric, others might be highly relevant depending on what other software you want to guarantee interoperability with.

I'm subscribed here to see how I can best assist. 🙂 Probably with ActivityPub experience more than with code. Feel free to tag me for questions that come up.

[Nachtpfoetchen]

so is that ment to be long term replacement of the existing @c3voc_releases@chaos.social bot? I think before deciding on the architecture a vision is needed what exactly is the goal. This is especially needed to account to following problems, like how to organize and handle the fediverse community as the moderation tools differ on what is available on youtube. Compared to youtube disabling replies doesn't exist on a protocal level (yet). So there needs a way to handle this, and especially who should handle this: The event organizers / event community managers vs the voc community managers? I think trying to get a vision how we want things like that to be handled will have great implications on how the architecture has to be designed to tackle this.

I think the current bot kinda works as there is not much interaction intended by an automated mastodon account, but thais will change if the presence of voctoweb is getting bigger.

[jfietkau]

On disabling replies: The GoToSocial project has a spec for interaction policies that is gaining momentum, which includes reply permissions. Those could be adopted by a custom server implementation, for example Fedify is tracking them at fedify-dev/fedify#453. In the meantime, while other software is still largely unaware of reply permissions, something that I do in one of my projects is to simply drop all incoming replies. Nothing says you have to host them if you don't want to.

[Haui1112]

so is that ment to be long term replacement of the existing @c3voc_releases@chaos.social bot? I think before deciding on the architecture a vision is needed what exactly is the goal. This is especially needed to account to following problems, like how to organize and handle the fediverse community as the moderation tools differ on what is available on youtube. Compared to youtube disabling replies doesn't exist on a protocal level (yet). So there needs a way to handle this, and especially who should handle this: The event organizers / event community managers vs the voc community managers? I think trying to get a vision how we want things like that to be handled will have great implications on how the architecture has to be designed to tackle this.

I think the current bot kinda works as there is not much interaction intended by an automated mastodon account, but thais will change if the presence of voctoweb is getting bigger.

Hi. I've asked last year about the implementation during 38c3. I work on different fedi implementations for some years now, such as peertube for kodi.

I think the big difference is that the fediverse (especially peertube) works as its own CDN. Videos can get federated and significantly lower the strain on the existing CDN. This also makes videos travel significantly wider outside the chaos universe. I think this is especially helpful for discovery and spread of information, which I think is a core value of the club.

I've been playing around with setting up a peertube instance and pointing it at media.ccc.de or the cdn. Peertube automatically allows the disabling of comments afaik.

Edit: This will not stop a video being commented on say mstdn.social but its not necessary to moderate that. Thats mstdn.social's problem.