OK folks, we need to stop assuming #Mastodon is the entire #Fediverse and expecting its idiosyncrasies to be mandated on everyone else.

Emelia 👸🏻

@silverpill @santisbon @julian
@by_caballero

I think there's two parts here: free-form content warnings where entropy is extremely high, and then more structured content labels, which should ideally have fairly low entropy.

(Cont)

Emelia 👸🏻

@silverpill @santisbon @julian @by_caballero

With content labels I was thinking along the lines of there being perhaps a few sets of them from well-known providers, e.g., from IFTAS, such that the URIs for a given label is the same no matter which AP software it came from, and supporting something like sameAs from the RDF/OWL world for linking tags from different providers.

I think if you do it per-community, you'll see fragmentation of the content labels, and they'll become less effective.

julian

@[email protected] if that's the case then I would caution you to keep the standard labels as vague as humanly possible, otherwise you'll end up playing referee over which specific, esoteric labels to add next (with very very vocal proponents, I'm sure!)

But... Why not both?

Possibly even a first line, standardized "opinionated" and "promotional" CW tags, and additional community specific CW tags.

Steve Bate

@santisbon @fbievan Nice diagram, but I don't see any Actor base type in the normative specs.

santisbon

@steve @fbievan they are called "actor types" so I had to come up with an abstraction for the diagram to cleanly convey their shared attributes.

Emelia 👸🏻

@silverpill @santisbon @julian @by_caballero

That's basically fine, in that there's pretty clear definitions for a core set like blood & gore, violence, politics, etc. but of course others can always define their own set of additional labels — the idea is to move the labels to be separate from the community, such that we reuse common labels, rather than needing to constantly block/filter additional copies lf effectively the same label

santisbon

@by_caballero @thisismissem I like the notion of a proper extension for a "hiding" feature. Since a CW and a spoiler alert are essentially the same mechanism it might be good to give it a generic name that doesn't exclude either use. Something like "Preview".

Used together with a sensitive attribute goes really well. For example, a news org would not hide the content for a politics post since that's exactly what its followers expect on their posts but might want to designate a specific image or video as sensitive.

Emelia 👸🏻

@santisbon @by_caballero so yeah, in what I'm envisioning, a platform about movies & tv could totally have a "Spoiler" label, perhaps many (one for each movie/tv show) and we could have a parent label that could be used to block/filter all child labels (linked data yay)

But you probably wouldn't create a label for each episode, and instead use the content warning free text for that

santisbon

@thisismissem @by_caballero the more I think about it the more I lean towards making it as simple and consistent as possible. I prefer the idea of a preview field so we can stop misusing summary i.e. use it consistently across Note and Article. If someone wants to hide the content it doesn't really make a difference if it's because it might spoil the movie or trigger someone's arachnophobia; the end result is the same. Then sensitive can be applied to things like images and videos, with or without the need for a preview. I'll keep giving it some more thought but that's where I am right now.

Steve Bate

@santisbon @fbievan I understand, but those /shared properties/ are wishful thinking. The AS2 core spec states "Actor objects are specializations of the base Object type" and they are defined "in only the most generalized way, stopping short of defining semantically specific properties for each." The AS2 vocab spec uses the phrase "Actor Types" but those are the specific types listed in the spec and those derive directly from Object (with no Actor-related abstraction).

Emelia 👸🏻

@santisbon @by_caballero yeah, I've given this a fair amount of thought, I think the "content warning" is what'll be most used, but I think some may opt into using structured labels too, which I think would be good, but only with finite labels

santisbon

@steve @fbievan since they are defined as "specific types of Actors" and contribute no additional fields to the concept of an Actor I found it useful to display it that way on my diagram for visualization purposes. It's weird to define it as literally a type of Actor when there's no Actor. Some examples where I would have gone in a different way from the spec:

- altitude should have Place as its domain, not Object.
- It's weird to have a specialization remove attributes from the entity it's extending the way collections and ordered collections do but alright.

Steve Bate

@santisbon @fbievan Again, I understand. After reading the AS2 spec numerous times, I think one must accept that the word "type" is used very loosely and informally in that context. Trying to make complete sense of the spec: "that way madness lies"... It's an interesting mental exercise though.

santisbon

@steve @fbievan maybe I'll add a note to the diagram to make it clear that the spec defines types of actors but no actor type and it's shown here for visualization purposes.

santisbon

@steve @fbievan on top of everything, since there are properties that are mandated for Actors in addition to those of Objects (like inbox) it makes it extra weird not to have an Actor type. So I find it useful to include it in the diagram, though as I said in another comment, maybe adding a note would help make it clear why it's there.

santisbon

@thisismissem @by_caballero I'd worry that devs would quickly feel too constrained by any set of labels and start extending them. I can't really argue against it in an extensible protocol, and before we know it we've reinvented the tag attribute or hashtags.

Mike Macgirvin 🖥️

Welcome to the fediverse. It isn't ruled by Twitter or Mastodon. We use hashtags to tag unsafe content here. It's part of the pre-existing culture you're trying to destroy in the quest to dominate and control. We also use Unicode and HTML. Length limits are optional. And we support using the 'summary' field as a summary.

Hashtags are content labels. What else did you think they were? They work across platforms and protocols and do not require an FEP. This solution wasn't chosen by geeks - it was chosen by the people of the fediverse long ago to tag their content. The original fediverse/federation projects built filters to let you choose what tags you wanted to collapse or block - if any. In this way the receiver is in control of what they don't wish to see. Not platforms, not protocols, not corporate owners. The sender just provides advice through the use of content tags (specifically hashtags). The platform (i.e. the fediverse or a specific project or product) is (or at least should be) neutral. It's a system that works.

silverpill

@mikedev Some developers need labels that are independent entities. Actors will use these (pre-defined) labels to categorize objects, including objects created by other actors.

Hashtags are usually ephemeral and are added by authors themselves.

For example Lemmy is working on a tag system but they don't want to use hashtags https://github.com/LemmyNet/rfcs/blob/main/0004-post-tags.md

Mike Macgirvin 🖥️

I've got no problem with community tags. We had those in Hubzilla years ago. The code is probably still there, but I don't think that project tries to federate them any more. That's a feature we lost in the move to ActivityPub as it never worked across projects and protocols. Whereas hashtags kept on working.

Which is my entire point.

Demon Queen Lucretia

@santisbon just for context, which common fediverse software *doesn't* have "content warning" fields or something closely aligned with that? This is my 4th fedi account now and everything I've used had that as a prominent feature.