What if... you had one Fedi account on a generic headless #ActivityPub server that simply hosts and federates your data... and had C2S UIs for microblogging, long form writing, media editing and sharing, link aggregation, games, fitness tracking, and so on, that all used that same Fedi account. Technically, it's a similar concept as ATProto (but no relay and app view) and Solid Pods (but no RDF).
It seems possible... if we can improve the AP C2S API/protocol sufficiently.
Steve Bate
🫧 socialcoding..
Evan Prodromou
flo
very normal personNote, it is artificially limited platform@steve This is exactly the sort of thing I'm hoping we see on the Fediverse
One thing that might need some careful design is: how to scope down the access that each of the C2S-connected apps have? A game for example doesn't need to be able to read all of my/my friends' messages, only game-related ones
I think the work Spritely's been doing could be a good building block for such a generic server
It depends.. since post-facto interop rules supreme, that's a possible approach to correct course.
I can't wait to see what you build! Good luck.
Maybe I am way over my head, but this seems like *exactly* what I am building right now and I'm not really building anything outside of ActivityPub C2S?
I mean... Yes, my current client assumes some specific profile for OAuth and the client will need a proxy to get some data remote servers (to bypass authorized fetch, or to resolve documents from transient activities), but doesn't seem to me that anything I am doing is outside of AP's scope?
Silverpill does not agree with us. I am wishing Silverpill well in their plan to rebuild with a completely different ground-up API. Silverpill is skilled and smart and very familiar with ActivityPub in great detail, so I'm sure they will come up with some interesting API architecture.
@raphael @silverpill @steve I don't intend to stop working on the ActivityPub API.
There are a lot of us collaborating through the SocialCG ActivityPub API Task Force, and I expect that collaboration to continue:
@evan @raphael @steve I think a generic ActivityPub server can't be built without major changes to ActivityPub. For example, it might be necessary to make all side-effects of activities explicit.
And there is a bigger problem. A generic server without FEP-ef61 is like ATProto PDS or Nostr relay but all data is tied to a single server. This means that it is obsolete before the work has even begun.
@steve @evan @raphael A server can't properly process an arbitrary activity without knowing its side effects. A server that only supports a small number of activities mentioned in the ActivityPub spec is obviously not generic.
I can point to other challenges because I've been working on this problem for years, but...
Several generic AP server implementations have been built
There are also simpler ways to design servers so that content isn't tied to a specific server
Wow, for real? I suppose it's time for me to retire then.
I don't follow: how does that relate to the "ActivityPub API" if the activity is "arbitrary" and not defined by ActivityPub, not using AS2 vocabulary?
Isn't that like saying that we can't use HTTP as a protocol because an HTTP server doesn't know what to do with verbs defined on, e.g, WebDAV?
Then I'd echo what @mariusor said: it seems like you are conflating the server with the applications built on top of it.
@silverpill @raphael @steve I think that comes down to what you mean by "support".
I think a server that delivers extension activities, or activities with extension properties or extension types as `object` etc , locally or remotely, is "supporting" those types.
Implementing server-side side effects of some extensions may be a bonus, but we can build a lot of great stuff without it.
@silverpill I think you're getting confused about ActivityPub side-effects and application logic side-effects.
I call my project FedBOX a "Generic ActivityPub server" because outside of storing ActivityPub objects and activities to a local storage and dispatching said activities to their recipients it doesn't do anything else.
However there's nothing preventing someone from forking the project and adding some other type of logic to it for specific combinations of Activities/Objects. That's the thing I'm trying to do with my GoActivityPub library: take care of the ActivityPub stuff, so you can then do your own stuff alongside it.
@silverpill what are extensions exactly? Are you talking about FEPs that proscribe behaviour alongside structure, or ActivityPub extensions as allowed by JSON-LD.
The FEPs that proscribe behaviour can't ever be done in a "generic" way, and I'm betting you already know that and you're just being facetious.
However the JSON-LD dynamic structure can be reasoned about generically from an ActivityPub point of view using the points I made above: storage to disk, dispatch to recipients.
If you plug smart clients on top of that that have the specific logic you want, you're all the way there to what Steve was talking about.
@mariusor @steve @evan @raphael An extension is a protocol extension. Specifically, I talked about activity types that are not mentioned in ActivityPub.
Here is a side effect of a Like activity:
The side effect of receiving this in an outbox is that the server SHOULD add the object to the actor's liked Collection.
Now imagine that you have an EmojiReact activity which server should add to object's emojiReactions collection as a side-effect.
A generic server should be able to do that.
I'm really not following this logic: you are saying "a generic HTTP server should be able to serve WebDAV requests".
Yes, HTTP servers should be able to serve WebDAV requests, *provided* they implement the extension.
It doesn't mean that they *have* to implement the extension. And an HTTP server that does not support the extension does not make non-HTTP compliant.
Jonathan Frederickson
Lutin Discret
silverpill
Raphael Lullis
mariusLike and EmojiReact are part of the same protocol, both are widely used in Fediverse. Not sure what is the point of twisting the meaning of words to make limitations of certain software appear as non-limitations.The analogy to HTTP is adequate. "GET/POST/DELETE" are verbs, PROPFIND is a verb. It just happens that some verbs had expectations around their behavior defined on spec, others are defined on an extension.
The same thing for AP: it says what to do about as:Like and as:Announce. If you want to attribute any special behavior for as:EmojiReact or for mitra:Subscribe, then it's up to the extension to define the behavior and get people to agree.
@silverpill @mariusor @evan @raphael "you have an EmojiReact activity which server should add to object's emojiReactions collection as a side-effect."
It's a direct rather than side effect, but how is that different from Add(object=EmojiReact, target=object_emojiReactions)? A generic server could support that.
@steve By side effect I mean anything that is not directly expressed in the activity, and EmojiReact usually doesn't specify that emojiReactions collection should be updated.
Add(object=EmojiReact, target=object_emojiReactions)? A generic server could support that.
Yes, this is what I said in the beginning of this sub-thread. All side-effects should be made explicit (at least for non-standard activities).
It is a generic server *because* it is focused only on the ActivityPub API, unlike most of the existing services that interop only with specific APIs (Mastodon's , Lemmy's).
A "generic server", in my view, is one that can take any type of activity posted to a box, and dispatch to the proper targets. The ActivityPub API defines only what to with like/announce/follow. Anything else is up to the application. That's the extensible part.
@silverpill @mariusor @steve @raphael
I think it's a fair critique and one that we should deal with.
However, I think it's possible to provide value where the server implements side effects of AP core, and not for any other Activity Vocabulary activities or extensions.
With the core AP activities, we can probably give clients most of what they need to implement "side effects" for extensions.
@silverpill @mariusor @steve @raphael
For example, let's say we decide that an `Arrive` activity should update the `location` property of the actor.
`Arrive` isn't a core activity in AP. So, instead of expecting the server to change the `location`, the client can send an additional `Update` activity to change the actor's `location` itself.
With CRUD activities and collection activities in core we can get a lot of this done from the client side.
@silverpill @mariusor @steve @raphael
I think there's value in having a negotiation between the client and the server so the server can declare its support for side effects for extensions. A way for the server to say, "I'll handle side effects for geosocial applications". Then the client can skip implementing those side effects itself.
@silverpill @mariusor @steve @raphael
More powerful would be a way for clients or actors to define side effect rules for servers dynamically. "If you get an Arrive activity, set the actor's location."
@silverpill @mariusor @steve @raphael
All that said, I think we can get pretty far with the core AP activities with defined side effects, and extension activities with "side effects" implemented by the client.
@naturzukunft2026 👋 I'm not exactly sure what you're asking, but there's a W3C ActivityPub API (C2S) Task Force meeting on March 19. That's where C2S extensions will be discussed. It's listed on the SocialCG events calendar.
Hey @evan ... I've looked into why I get issues with checkin.swf.pub, and it could be that I'm doing something wrong, but it hits my servers using client_id=https://checkin.swf.pub/client.jsonld
Which, when gets dereferenced serves an ActivityPub Actor(ish) like document instead of the Client ID Metadata I'm expecting.
I remember I was arguing with Emelia about how to server both formats from the same URL, but in the end I decided it's not worth my time trying to do both, and I'm just expecting the data model described in the RFC7591 Sect2: https://datatracker.ietf.org/doc/html/rfc7591#section-2
I love this vision. One hurdle I see is the lack of content types + the ability to represent them in HTML.
Bandwagon is my best use case: Albums and Tracks aren’t a standard data type. I could push those into my own AP server, but other servers wouldn’t know how to represent them correctly.
For this to work, we need a restricted subset of HTML for generating previews of unknown content.
This feature exists in oEmbed (optional) but is poorly supported on the Fediverse.
@benpate as long as your custom types have the current ActivityStreams Object as a base (ie, they contain a Content and a MediaType) anyone would be able to render them to some extent.
For cases where you have a different structure, you can have Profile objects alognside them linked to their Preview property. Use the vocabulary to your advantage.
naturzukunft
fredy
KungFuDiscoMonkey
Matthias Pfefferle
Lorraine Lee
Ben Pate 🤘🏻