Rob Ricci
@ricci.io
Research Professor in the Kahlert School of Computing at the University of Utah
Not planning to be here much, my main account is on the fediverse at https://discuss.systems/@ricci/ - you can follow it directly here as
@ricci.discuss.systems.ap.brid.gy
Not planning to be here much, my main account is on the fediverse at https://discuss.systems/@ricci/ - you can follow it directly here as
@ricci.discuss.systems.ap.brid.gy
I was very curious about this, so I sampled 100,000 dids that have been active in the last week. The number that appear to have rotation keys of their own (replacing or in addition to the keys managed by their PDS) is 8.
Still collecting more data, will post more detail when I have it.
Still collecting more data, will post more detail when I have it.
November 30, 2025 at 7:32 AM
I was very curious about this, so I sampled 100,000 dids that have been active in the last week. The number that appear to have rotation keys of their own (replacing or in addition to the keys managed by their PDS) is 8.
Still collecting more data, will post more detail when I have it.
Still collecting more data, will post more detail when I have it.
And this is a way to help move the network - maybe - in a direction that aligns its deployment and use with its ideological principles.
November 29, 2025 at 8:24 PM
And this is a way to help move the network - maybe - in a direction that aligns its deployment and use with its ideological principles.
The second would require PDSes to expose information about whether accounts are downloading CAR files, so it's probably unanswerable in practice at least for the time being.
November 29, 2025 at 8:09 PM
The second would require PDSes to expose information about whether accounts are downloading CAR files, so it's probably unanswerable in practice at least for the time being.
I suspect that the first half of this question is answerable, at least heuristically, today, by looking at audit logs in the PLC. Maybe if someone hasn't already done it, I'll give it a go.
November 29, 2025 at 8:09 PM
I suspect that the first half of this question is answerable, at least heuristically, today, by looking at audit logs in the PLC. Maybe if someone hasn't already done it, I'll give it a go.
This is why it's worth measuring: the measurements are not ends unto themselves, they are ways of understanding how the network has been built in practice and how it's being used, and therefore getting past the dogma of the protocol design and into the practical deployment.
November 29, 2025 at 8:04 PM
This is why it's worth measuring: the measurements are not ends unto themselves, they are ways of understanding how the network has been built in practice and how it's being used, and therefore getting past the dogma of the protocol design and into the practical deployment.
And I want to make sure my point is 100% clear here: PDS migration is a fantastic feature of atproto, and it's great that the bluesky team planned for adversarial migration right from the beginning. My point is that the network is more than just the protocol, it's how people use the protocol.
November 29, 2025 at 8:04 PM
And I want to make sure my point is 100% clear here: PDS migration is a fantastic feature of atproto, and it's great that the bluesky team planned for adversarial migration right from the beginning. My point is that the network is more than just the protocol, it's how people use the protocol.
The point is that if you use the same account everywhere, you are giving one entity the ability to cut off that account everywhere
November 29, 2025 at 7:18 PM
The point is that if you use the same account everywhere, you are giving one entity the ability to cut off that account everywhere
BTW, I was incorrect about the number, now that I am at my computer I can actually check, and there are ~120 active accounts on the tngl dot sh PDS
November 29, 2025 at 7:16 PM
BTW, I was incorrect about the number, now that I am at my computer I can actually check, and there are ~120 active accounts on the tngl dot sh PDS
The point is simply that there are big risks in having the same entity be able to take down both your social media posts and the account that control access to your code. And the more eggs you put in this basket, the bigger the risk.
Many people will of course gladly accept this risk
Many people will of course gladly accept this risk
November 29, 2025 at 7:15 PM
The point is simply that there are big risks in having the same entity be able to take down both your social media posts and the account that control access to your code. And the more eggs you put in this basket, the bigger the risk.
Many people will of course gladly accept this risk
Many people will of course gladly accept this risk
They are, and there are around a dozen users on it. If you use their PDS, you are doing so with a different identity than the one you use for Bluesky etc., so it's a separate account and you're not getting much of the benefit of the architecture
November 29, 2025 at 6:43 PM
They are, and there are around a dozen users on it. If you use their PDS, you are doing so with a different identity than the one you use for Bluesky etc., so it's a separate account and you're not getting much of the benefit of the architecture
That certainly seems to be what BlueSky is trying with stacked moderation. So far, it doesn't look to me like that's seen as a killer feature for most users here, but maybe we just need to give it time
November 29, 2025 at 6:40 PM
That certainly seems to be what BlueSky is trying with stacked moderation. So far, it doesn't look to me like that's seen as a killer feature for most users here, but maybe we just need to give it time
and similarly if you wanted the "complete view of the world" design atproto has, ActivityPub ain't never going to scale to that. So you can approximate features of one in the other, but it's unlikely you'd be able to build networks that have the same shape in terms of who has control over what
November 29, 2025 at 6:38 PM
and similarly if you wanted the "complete view of the world" design atproto has, ActivityPub ain't never going to scale to that. So you can approximate features of one in the other, but it's unlikely you'd be able to build networks that have the same shape in terms of who has control over what
I don't think you can directly build either on the other protocol: while relays are easier to build with sync 1.1, the network is not going to scale to the 32k relays you'd need to get the same degree of individual community control over data and visibility that the fediverse has ...
November 29, 2025 at 6:38 PM
I don't think you can directly build either on the other protocol: while relays are easier to build with sync 1.1, the network is not going to scale to the 32k relays you'd need to get the same degree of individual community control over data and visibility that the fediverse has ...
As a standard it predates atproto but I'm not sure when atproto committed to supporting it. Last I checked there were a few dozen accounts using it, so pretty rare
November 29, 2025 at 6:30 PM
As a standard it predates atproto but I'm not sure when atproto committed to supporting it. Last I checked there were a few dozen accounts using it, so pretty rare
If you are interested in the raw data I'm building the charts from, let me know and I will work out a way to get it regularly uploaded to the site
November 28, 2025 at 6:28 PM
If you are interested in the raw data I'm building the charts from, let me know and I will work out a way to get it regularly uploaded to the site
My feeling is that what is needed is (a) some of ☝️and (b) smaller communities that set and enforce their own standards.
You might be able to guess where I'm going with this.
You might be able to guess where I'm going with this.
November 28, 2025 at 6:28 PM
My feeling is that what is needed is (a) some of ☝️and (b) smaller communities that set and enforce their own standards.
You might be able to guess where I'm going with this.
You might be able to guess where I'm going with this.
this feed will show you: bsky.app/profile/did:...
The dominant use of non-bsky PDSes right now, in terms of numbers, FWIW, is bridging in accounts from other social networks.
The dominant use of non-bsky PDSes right now, in terms of numbers, FWIW, is bridging in accounts from other social networks.
November 28, 2025 at 6:27 PM
this feed will show you: bsky.app/profile/did:...
The dominant use of non-bsky PDSes right now, in terms of numbers, FWIW, is bridging in accounts from other social networks.
The dominant use of non-bsky PDSes right now, in terms of numbers, FWIW, is bridging in accounts from other social networks.
I'm one of them! :) Well, the own PDS thing, I wanted to do did:web but I already had an established account when I learned about it, and you can't migrate since it's a different identity. If you want to see what your timeline looks like with only people who are not on the bluesky PDSes ...
November 28, 2025 at 6:27 PM
I'm one of them! :) Well, the own PDS thing, I wanted to do did:web but I already had an established account when I learned about it, and you can't migrate since it's a different identity. If you want to see what your timeline looks like with only people who are not on the bluesky PDSes ...
Yep, agreed. That's why I think it's worth watching the numbers here, because it started from a place of single code base, let's see if it grows significantly past that (yes I know about rsky) or whether the same forces dominate
November 28, 2025 at 4:04 PM
Yep, agreed. That's why I think it's worth watching the numbers here, because it started from a place of single code base, let's see if it grows significantly past that (yes I know about rsky) or whether the same forces dominate
(I assume you know the drama here that I'm referencing, the article is here: www.techdirt.com/2022/09/23/t...) just goes to show how correct that axiom is
The Most Famous Blunder Of Content Moderation: Do NOT Quote The Princess Bride
We’ve written stories about people having difficulty recognizing people joking around quoting movies. Sometimes it ends up ridiculously, like the guy who was arrested for quoting Fight Club a…
www.techdirt.com
November 28, 2025 at 3:45 PM
(I assume you know the drama here that I'm referencing, the article is here: www.techdirt.com/2022/09/23/t...) just goes to show how correct that axiom is
Yeah, I found it a funny irony that one of the recent moderation dramas here was extremely similar to one that Mike wrote a blog post about a few years ago (Johnny Cash lyrics vs. Princess Bride quote)
November 28, 2025 at 3:40 PM
Yeah, I found it a funny irony that one of the recent moderation dramas here was extremely similar to one that Mike wrote a blog post about a few years ago (Johnny Cash lyrics vs. Princess Bride quote)