Why did #ActivityPub create special behaviors such as Like, Announce and Block (and the Undo variants) instead of using Add to or Remove from the associated Collection objects?
-
@eyeinthesky @smallcircles and how do I infer social meaning from you adding an object to a collection? I could only do that if I was adding an activity to a collection
@thisismissem @smallcircles I’m not sure how you define “social meaning”, but if you Add a Boop to my boops collection you socially booped me. I suppose a Boop activity is needed either way, but overloading the meaning to include implicit collection management seems dubious to me.
-
@eyeinthesky @smallcircles that's the thing though! The likes collection doesn't contain activities, it contains objects.
@thisismissem @smallcircles Per the AP spec… “The side effect of receiving this in an inbox is that the server SHOULD increment the object's count of likes by adding the received **activity** to the likes collection if this collection is present.” (emphasis mine)
-
@eyeinthesky @smallcircles (funnily enough, in AT Protocol, we only have collections we add records to, all the social layer is defined by applications above, which receive all the changes to records and infer meaning from that)
I just responded in the other thread about #ForgeFed way of dealing with Issues and PR's, emphasizing the actor model in the #ActivityPub specs.
Guess it depends on the nature of your extension and its desired functionality how you model things, but ForgeFed chose to have a dedicated actor, a TicketTracker, to manage the collection. Big advantage is that it become a truly encapsulated service with its own business logic, fronted by an AP actor for interoperable network communication.
🫧 socialcoding.. (@smallcircles@social.coop)
@thisismissem@hachyderm.io @deadsuperhero@social.wedistribute.org I was looking at #ForgeFed which is a very sizable #ActivityPub extension (constituting the "Code forge" app domain in app-centric view, but arguably "Software development" top-level business domain in a service-oriented fedi). The way that things are modeled here adheres more to the actor model where there's a Factory actor, which in turn creates resource actors that expose various sub-domains. For instance for the management of Issues and PR's there's a TicketTracker actor to obtain via a Factory actor on a forge instance. Though I'm not sure whether I'd modeled that in similar fashion, it is a fascinating direction where we focus much more on good protocol extension design. All in all AS/AP offers a very granular foundation that allows for very interesting architectures, if only we dare explore them and do not dogmatically stick to some engrained notion how "social media" ought to be. I see #SocialMedia as but a small subset of #SocialNetworking.
social.coop (social.coop)
-
@thisismissem @smallcircles I’m not sure how you define “social meaning”, but if you Add a Boop to my boops collection you socially booped me. I suppose a Boop activity is needed either way, but overloading the meaning to include implicit collection management seems dubious to me.
@eyeinthesky @smallcircles yeah, and this is a side effect of the activity (e.g., Like) adding to both my likes collection and the likes collection on your post.
So would my add be targeting two collections with an Activity?
What about Announce? That's add to my outbox, add to my boosts collection, and add to your post's boost collection.
-
@thisismissem @smallcircles Per the AP spec… “The side effect of receiving this in an inbox is that the server SHOULD increment the object's count of likes by adding the received **activity** to the likes collection if this collection is present.” (emphasis mine)
@eyeinthesky @smallcircles I suspect this is incorrect. I am pretty sure the likes collection contains only objects, not the activities.
@evan was there an erratum for this? I feel like there was?
-
I just responded in the other thread about #ForgeFed way of dealing with Issues and PR's, emphasizing the actor model in the #ActivityPub specs.
Guess it depends on the nature of your extension and its desired functionality how you model things, but ForgeFed chose to have a dedicated actor, a TicketTracker, to manage the collection. Big advantage is that it become a truly encapsulated service with its own business logic, fronted by an AP actor for interoperable network communication.
🫧 socialcoding.. (@smallcircles@social.coop)
@thisismissem@hachyderm.io @deadsuperhero@social.wedistribute.org I was looking at #ForgeFed which is a very sizable #ActivityPub extension (constituting the "Code forge" app domain in app-centric view, but arguably "Software development" top-level business domain in a service-oriented fedi). The way that things are modeled here adheres more to the actor model where there's a Factory actor, which in turn creates resource actors that expose various sub-domains. For instance for the management of Issues and PR's there's a TicketTracker actor to obtain via a Factory actor on a forge instance. Though I'm not sure whether I'd modeled that in similar fashion, it is a fascinating direction where we focus much more on good protocol extension design. All in all AS/AP offers a very granular foundation that allows for very interesting architectures, if only we dare explore them and do not dogmatically stick to some engrained notion how "social media" ought to be. I see #SocialMedia as but a small subset of #SocialNetworking.
social.coop (social.coop)
@smallcircles @eyeinthesky yeah, making everything an Actor is certainly a choice, and you could do that
-
@eyeinthesky @smallcircles I suspect this is incorrect. I am pretty sure the likes collection contains only objects, not the activities.
@evan was there an erratum for this? I feel like there was?
@thisismissem @smallcircles @evan It makes sense to me. Someone liked a local object. We wouldn’t add the liked object to the liked object “likes” collection (it would be the same object). It seems that only the activity (and collection count) is interesting. Maybe you are thinking about the “liked” collection? (Different topic, but it seems that this *should* be a collection of activities too. Dropping the activity loses info such as the Like timestamp.)
-
@smallcircles @eyeinthesky yeah, making everything an Actor is certainly a choice, and you could do that
I think we underrated the power of the actor model and the extent we can incorporate it on the #ActivityPub fediverse. Somehow we got eternally stuck in talking about HTTP plumbing and core protocol capabilities that we never fleshed out thoroughly in order to be able to just focus on the higher-level concerns of app and service modeling.
Actor systems based on loosely-coupled event-driven architecture, delegation, supervision, supervision strategies, inbox strategies, let-it-fail, actors fronting domain aggregates, service-orientation, etc.
-
I think we underrated the power of the actor model and the extent we can incorporate it on the #ActivityPub fediverse. Somehow we got eternally stuck in talking about HTTP plumbing and core protocol capabilities that we never fleshed out thoroughly in order to be able to just focus on the higher-level concerns of app and service modeling.
Actor systems based on loosely-coupled event-driven architecture, delegation, supervision, supervision strategies, inbox strategies, let-it-fail, actors fronting domain aggregates, service-orientation, etc.
The biggest folly imho is this idea of "let's cram every domain into #ActivityStreams somehow". Flatten everything and project it onto this small set of social primitives that AS defines.
It is once more a choice of pragmatism: "Hey, I've seen it working with Mastodon, so I copied that. And #LinkedData extension mechanism is a handwaved horror show".
So understandable perhaps that we did it. But now we must overcome this trend which has taken stubborn root and drags the ecosystem down.
-
@silverpill@mitra.social add and remove feel more like building blocks than actual activities.
It would never capture the nuance of many of the activities that it would purportedly represent.
-
@silverpill@mitra.social add and remove feel more like building blocks than actual activities.
It would never capture the nuance of many of the activities that it would purportedly represent.
@julian @silverpill Add/Remove don’t replace the activity being added/removed so there’s no risk to nuance from my perspective. They make the collection side-effects explicit and consistent with other non-special collections. Sometimes the Add (or Remove) is the primary activity rather than only a building block (Add a member to my Group, Add an Article to my fave articles, Add a song to a social play list — maybe with an Announce of the Add sent to followers).
-
@eyeinthesky my assumption for why there's multiple methods for achieving relatively similar results is that first the social vocabulary was created as a way to express the actions that the existing platforms already had enshrined, and then other operations, like collection management, were needed.
But as @thisismissem already said, the canonical social activities have specific side effects attached to their behaviour that extend to the entire social graph, and those are what ActivityPub actually codifies into its specification.
@mariusor
The earliest discussions on Activity Streams did constantly get stuck on which platforms' concepts were "real" and which was just an alias for another.e.g. is a "tweet" the same thing as a blog post? As a status update on Facebook?
e.g. is listening to a podcast the same verb as watching a video on YouTube?
e.g. is "liking" the same verb as adding to a "Favorites" playlist?In retrospect I think it was naive to try to be so specific, since it was too early for that, but alas.
️ -
@mariusor
The earliest discussions on Activity Streams did constantly get stuck on which platforms' concepts were "real" and which was just an alias for another.e.g. is a "tweet" the same thing as a blog post? As a status update on Facebook?
e.g. is listening to a podcast the same verb as watching a video on YouTube?
e.g. is "liking" the same verb as adding to a "Favorites" playlist?In retrospect I think it was naive to try to be so specific, since it was too early for that, but alas.
️@apparentlymart very interesting perspective, thank you.
-
The biggest folly imho is this idea of "let's cram every domain into #ActivityStreams somehow". Flatten everything and project it onto this small set of social primitives that AS defines.
It is once more a choice of pragmatism: "Hey, I've seen it working with Mastodon, so I copied that. And #LinkedData extension mechanism is a handwaved horror show".
So understandable perhaps that we did it. But now we must overcome this trend which has taken stubborn root and drags the ecosystem down.
@smallcircles @eyeinthesky I was mostly thinking of the Follow activity and the following collection: it doesn't contain Follow activities, it contains Actors: https://hachyderm.io/users/thisismissem/following?page=1
In Mastodon the "likes" collection for a Note doesn't even include the Actors, just the count: https://browser.pub/https%3A%2F%2Fhachyderm.io%2Fusers%2Fthisismissem%2Fstatuses%2F115622304738688291%2Flikes
-
@thisismissem @smallcircles The weird thing about Like+side_effects vs Add(Like activity to “likes collections”) is that the side effect is Add Like to “likes collection”. lol They are both one inbox activity.
@eyeinthesky @thisismissem @smallcircles i would use as:result for this -- the *result* when you Accept the Like activity is to Add it to the likes collection... if you care about that level of detail. most people only care about the Like. maybe even less than that! (synthesizing a statement such as :bob :likes :this.)
-
@eyeinthesky @smallcircles that's the thing though! The likes collection doesn't contain activities, it contains objects.
@thisismissem @eyeinthesky @smallcircles likes contains Like activities; shares contains Announce activities; the exception is followers not containing the Follow activity (or not having a follows collection) (or not doing the better thing which is subscription management)
-
@silverpill @eyeinthesky @thisismissem @smallcircles the activity properties don't get used very much but they probably should be. the as:actor performs the as:Activity on the as:object to the as:target from the as:origin with the as:instrument leading to the as:result. and maybe in response to the as:inReplyTo? maybe with an as:summary? there are properties of as:Object that would fit well on most as:Activity nodes, although i think they are two distinct processing models (just closely related)
-
@silverpill @eyeinthesky @thisismissem @smallcircles the activity properties don't get used very much but they probably should be. the as:actor performs the as:Activity on the as:object to the as:target from the as:origin with the as:instrument leading to the as:result. and maybe in response to the as:inReplyTo? maybe with an as:summary? there are properties of as:Object that would fit well on most as:Activity nodes, although i think they are two distinct processing models (just closely related)
@silverpill @eyeinthesky @thisismissem @smallcircles which is to say, as2 has an "object processing" mode (check name, summary, content, icon, image, ... and render generic display) and an "activity processing" mode (check actor, object, ... and render an activity in the stream) which coexist. the "activity processing" is underutilized in fedi, but it's the more interesting one because it lets you unify a lot of disparate models without forcing everything to be a Note (checkins, scrobbles, etc)
-
R relay@relay.an.exchange shared this topic
