Are activities without actors allowed?

[snarfed]

AS2 says all properties are optional, including id and actor. AP confirms that id is common but not strictly required, eg for transient activities. Is actor also optional in activities? (Thinking about S2S primarily.) Is there a use case for actor-less activities? Are they maybe theoretically possible, but in practice pretty much everything will discard and ignore them? Is this worth guidance anywhere?

(There's obviously one conversation direction here around authn, checking HTTP signature keyIds against actors, etc, and most fediverse software will likely fail actor-less activities early due to that, but I'm more interested in the semantics and expectations of actor-less activities, if any, not authn/authz.)

[snarfed]

Damn, meant to file this in https://github.com/w3c/activitypub/issues . Feel free to move there if appropriate.

[evanp]

So, I think it does make sense to move this to AP. AS2 is extremely relaxed about properties, and you are correct, it's perfectly reasonable to have an Activity type object with no actor property. What that means in a social networking environment or federated system is another topic. Let's move this to AP, and if we think there are general rules to follow, we can copy it back here eventually.

[evanp]

I think this is a really interesting question. If you think of AS2 activities as sentences (not that crazy!), then the kind of sentence in English that don't have a very well-defined subject are sentences that start with "There is" or "It is".

• "There is a person named Ryan." This kind of "existence activity" isn't one we deal with often, but interesting to call out.
• "It is 65 degrees in San Francisco today". There isn't a clear actor in measurement or environmental conditions, but these are the kind of activities that we see in social network quite a lot.

So, I think there are some options for having activities that aren't anchored to an Actor, but the question then becomes: where do they live? They aren't in anyone's "outbox" for example.

I don't have an easy answer here; probably something to think about for the next version or a FEP.

[github-actions]

This issue has been labelled as potentially needing a FEP, and contributors are welcome to submit a FEP on the topic.
Note that issues may be closed without the FEP being created; that does not mean that the FEP is no longer needed.

[snarfed]

Thanks!

@evanp's "existence activities" are a good example use case. To flesh them out, say you want to inform a remote server of an object that you don't own or control. https://places.pub/ is a great example. It might send out Creates for its place objects, eg from a server/instance actor, but it can't send them everywhere. Say indieweb.social doesn't know of a places.pub place, and I want to inform it of that place. I don't own the place object, but I could send something like this:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "type": "Create",
  "object": "https://places.pub/relation/1224652"
}

indieweb.social could see this, fetch https://places.pub/relation/1224652, and cache and use it locally. Theoretically the Create doesn't need an actor or authn/authz.

[trwnh]

{
  "type": "Activity",
  "summary": "An activity happened"
}

{
  "type": "Create",
  "object": {"name": "Something"},
  "summary": "Something was created"
}

{
  "actor": {"name": "Ryan"},
  "type": "Create",
  "object": {"name": "Something"},
  "summary": "Ryan created Something"
}

you are free to selectively state or disclose as much or as little as you want or need.

---

there is also the idea that the actor may be inferred from other information (correctly or incorrectly). for example, i'm pretty sure i saw some (relay?) software use the keyId= of an http Signature and ignore the actor in the body content. the reasoning there seems to be that the actor who signs for the delivery should be the actor who performed the activity in all cases.