Why am I here and not on #Bluesky or #Threads?

Steve Bate

@jwildeboer Just Just FYI, your server is configured to publicly expose your following/followers connections. Harvesting exposed metadata across a federated network is not really very expensive.

Raphael Lullis

@jwildeboer

Wait, this argument does not hold.

Like you said, the data is already public on ActivityPub. Whatever adversaries interested in building the social graph or analyzing metadata can do so, regardless of your instance location or who is in power.

Jan Wildeboer 😷:krulorange:

@pravee_n See my other thread at https://social.wildeboer.net/@jwildeboer/113487613965056474 where I explain why Bluesky isn't decentralised and why running your own PDS means nothing as they still run other essential parts. @heaths

fury999io

@jwildeboer Why specifically Trump?

Jan Wildeboer 😷:krulorange:

@fury999io Becaise he is becoming president of the country where Bluesky, Facebook etc are based.

nelomah

@raphael @jwildeboer does ActivityPub include metadata?

Jan Wildeboer 😷:krulorange:

@raphael They’d have to check/scrape *all* ActivityPub instances all the time to get a complete social graph of the Fediverse. While at a centralised service that social graph is part of the architecture. That’s a huge difference.

Raphael Lullis

@nelomah @jwildeboer

How could it not?

Jan Wildeboer 😷:krulorange:

@raphael Or to explain my argument in more simple ways: Twitter has the Firehose, where you can access all traffic at once. In a federated network such a firehose doesn’t exist and trying to build it is an exponentially bigger problem that IMHO cannot be solved.

Raphael Lullis

@jwildeboer

Oh, please... I was toying with the idea of making a search engine for the Fediverse last year. It took me less than a week to scrape and index 8 million ActivityPub accounts last year, with all of the users posts, including those that had enabled authorized fetch.

From my home computer.

On a shitty 50MB DSL connection from O2.

Raphael Lullis

@jwildeboer

Capital-P "privacy" is absolutely incompatible with publishing networks. Telling people that Mastodon is any better in this regard is at best wishful thinking and at worst irresponsible.

We *might* make it work with AP if we have better C2S implementations where end users control the keys and allow for E2EE, but if your threat model involves 3LAs and corporations building your profile, the best solution is to avoid any social network and stick with Signal/Matrix/XMPP.

Jan Wildeboer 😷:krulorange:

@raphael I specifically pointed at that in the second toot on my thread: „I still treat every single post and message here as ultimately being stored and analysed by adversaries because for me the Pub in ActivityPub means public.“

Raphael Lullis

@jwildeboer

Okay, cool. My objection though is to the first post. It doesn't matter whether we are here, Threads, Bluesky or Twitter. When it comes to social media, the only winning move for those concerned about state actors and large corporations is not to play.

Jan Wildeboer 😷:krulorange:

Now also part of a blogpost at https://jan.wildeboer.net/2024/11/My-Digital-Life-After-Twitter/

Heath Stewart

@jwildeboer @pravee_n my understanding was that anyone *could* run a relay, but even their docs admit it’s a ton of bandwidth and implies there’s not much incentive to do so. So maybe they are practically running all the relays and that makes them practically centralized.

Still, it seems to solve for the problem of the Mastodon effect where sites are getting DOS’d by instances requesting data for thumbnails.

Personally, I see pros and cons. To each their own, though.

Jan Wildeboer 😷:krulorange:

@heaths To identify yourself in the Bluesky network, you need an identifier. The layer that manages and provides these is the PLC. And, guess, who runs the only PLC that is used in the Bluesky network?@pravee_n

praveen

@jwildeboer @heaths Is it possible to tun alternative PLC?

Jan Wildeboer 😷:krulorange:

@pravee_n AFAICS no. @heaths

Heath Stewart

@jwildeboer @pravee_n looks like you can use the `did:web` method as described in https://atproto.com/specs/did#blessed-did-methods

To clarify: not trying to push anyone anywhere. This just seems like a great exploration of the protocol. I skimmed the docs earlier - I’m really liking the Labeler functionality - but this thread has convinced me to read more thoroughly.

Jan Wildeboer 😷:krulorange:

@heaths @pravee_n No, you can’t. did:plc is primary. And where do you store the accepted did:web? Right. In the PLC layer that is owned/run by Bluesky. This is a pattern, BTW. Whatever *could* be decentralised at some point in the future „unfortunately“ right now is Bluesky operated.