Other people in that thread have pointed out that it isn’t showing posts being delivered to Threads despite the block. That should be testable with other instances, but not Threads since it’s not receiving any content from Mastodon at the moment. The concerning thing there is the user still being able to view content from people they’ve blocked, but that seems to be a bug if it’s reproducible.

In the EU companies can’t scrape personally identifiable information without consent, even if it’s already publicly available. IANAL, and there’s probably ways they can sneak around the GDPR, but at least it’s not a free for all. It’s unclear though how it works for federation. It’s definitely not the same legally though.

The reason for not directly federating content to Threads isn’t so nobody there can ever see my amazing posts, it’s so Meta can’t easily profile me. Scraping public posts on a different platform would probably be illegal, at least in the EU, and reposts don’t give them a lot of data about me. Federating content, however, would give them most of the same data that Mastodon has on me without even having to ask.

This post from Eugen Rochko mentions that blocking Threads at the user level “stops your posts from being delivered to or fetched by Threads”. Basically, the user-level instance block is bidirectional.

Limited federation mode is a different feature, at the admin level. It doesn’t really affect the delivery of posts in either direction, it just hides the blocked instance’s content from the global feed. Defederation on the other hand is indeed bidirectional, but again it’s on the admin level rather than users’.

Mastodon instance blocks are already bidirectional AFAIK: if you block an instance your content does not get federated with them. I was actually surprised that this does not seem to be the case for Lemmy. I don’t think this break any core abstraction of AP…

Interesting demo! Does this use the user agent string for identifying clients?

Oh I mean, sure, but I don’t think IP logging is the main privacy concern with spy pixels.

I’m assuming this trick uses the user agent string and other request metadata to identify clients. Even if it didn’t recognize Jerboa as a client, it did guess that I was on mobile. That’s not possible just by tracking IPs, unless they’re cross-referencing it with other datasets. Also, I was on VPN anyway, so the IP would have been useless.

It should be possible for clients to obfuscate/fake the metadata of image requests to make tracking with spy pixels less effective.

Can countermeasures be implemented in the clients to mitigate privacy risks, while not having to proxy images?

Awesome work folks, thanks!