shrugal

Yea, but that's just a lowest common denominator (e.g. it doesn't include things like lemmy community sidebars), and also generally not appropriate for a client application. ActivityPub transmitts all events that are happening (posts, likes ...) between servers, and they are supposed to index and aggregate things (e.g. sum up votes, sort posts). It's just not feasible to expect the same from a mobile app for example, you'd have to at least create another standard for that.

So services end up implementing their own client APIs to fit their needs. And imo that's actually a good thing, because it allows them to try out features and specialize on different use cases. But afaik the ActivityPub people are working on another standard for client APIs, at least it's on their radar.

Because they are still different apps with different needs, architectures and formats. They just synchronize most of their content between each other.

That's what I thought as well. If the authors of this "study" were able to simply scan for it on the Fediverse, then what's stopping law enforcement units from doing the same? They can literally get a message everytime someone posts something on a suspicious instance.

Same. But I also worked as a dev for an online advertising company, and conversions are everything to them! If this causes 5% more users to not close the tab because of a captcha check, then every last one of them will want to have this.

Although if there was an alternative and easier way to prove that I'm human on the internet, without harming my privacy or allowing someone to arbitrarily block legitimate users (like this proposal), then I would be all for it. The problem here is that the checking standards and process would be in the hands of a few companies, so they could check for much more than just that.

I think your and their definition of "trusted" is a bit different. They mean trusted as in "very likely a real human". That's not enough to allow any privileged access, but it should help when trying to block bots heuristically while preserving a good experience for real users. "Trusted" devices could skip capture checks for example.

Of course this doesn't make this proposal any better, it's still extremely dangerous and misguided imo!

I guess something like: Skip capture if it succeeds, show capture if it fails. It would allow people to skip capture checks most of the time.

To be clear, this doesn't make it ok!

The study doesn't compare their findings to any other platform, so we can't really tell if those numbers are good or bad. They just state the absolute numbers, without really going into to much detail about their searching process. So no, you can't draw the conclusion that the Fediverse has a CSAM problem, at least not from this study.

Of course that makes you wonder why they bothered to publish such a lackluster and alarmistic study.

Why do they just mention absolute numbers, instead of comparing them to similar platforms? All they said was that there is CSAM on the Fediverse, but that's also true for centralized services and the internet as a whole. The important question is whether there is more or less CSAM on the Fediverse, no?

This makes it look very unscientific to me. The Fediverse might have a CSAM problem, but you wouldn't know it from this study.

Afaik the J series is known to have bad performance when it comes to processing media, like creating thumbnails or transcoding videos. That's why they added the Play series at some point, as an affordable option with ok media support.