Hacker News Reader

Comment by ekosz | Hacker News Reader

Reading articles like this one, I do think people are trying to put a square peg (their applications) through a round hole (ATProto). The ATProto was designed around all data being public. You write public data to a user's PDS and then any application can read that public data and do something with it. If all data was private (encrypted?) by default then that would defeat half of all of ATProto's goals.

Imagine someone starts startup A on-top of the ATProto. They raise some money, get some users, some people love it, but ultimately they die. If the data was private to that service, that data dies with the startup. But if the data is all public future startup B can read that old data and do something with it. As a user that's brings me a ton of utility and comfort trying out new services.

If you're trying to built a local-first, mostly private service I just don't think the ATProto is the right tool for the job.

View on HN

> If you're trying to built a local-first, mostly private service I just don't think the ATProto is the right tool for the job.

I agree! I think the "permissioned data" working group is aiming to solve the middle ground – where you want to broadcast something "publicly" but to a specific audience instead of the whole world (e.g. invite-only event, membership club). It's "private" in the sense it's not open to all but not in the sense that only you can see the data.

Nick Gerakines has been sharing some good examples of what you could build using permissioned data:

- Private Events: https://ngerakines.leaflet.pub/3mqxalpvn4k2e

- Bookmarks: https://ngerakines.leaflet.pub/3mqu653us3k2p

- Community content: https://ngerakines.leaflet.pub/3mqzsstcsok25

- Forums: https://ngerakines.leaflet.pub/3mr3uqjjaxs2d

- Polls: https://ngerakines.leaflet.pub/3mr3waxevjc24

atproto permissioned data [1] will solve the dilemma you're describing. Ultimately all the data is in the same place (your PDS) and you can always read and write any data there. If startup A goes away, startup B can read that old data and do something useful with it.

[1] https://github.com/bluesky-social/proposals/blob/main/0016-p...

> If all data was private (encrypted?) by default then that would defeat half of all of ATProto's goals.

Yeah, this is the reason why I don't understand why they succumbed to the idea ATProto must handle private and has started on work trying to figure it out (https://atproto.wiki/en/working-groups/private-data). Instead, focus on just really great public data archiving and displaying, at scale.

> I don't understand why they succumbed to the idea ATProto must handle private

The answer is very simple, the people demand it.

Technically, or if you squint the right way, it is a new protocol (atp://) that shares some parts with the public side. Notably the relay is out (for now?) and the data lives in a different sqlite table (iirc/aiui).

Yeah, running a social website where you can't write anything in private to just your friends is not very based.

Like this one?

HN is an exception to this rule imo, and remains our diamond in the rough

Are you saying you are here to make friends on this thing?

If the primary user concern is "all data is public", then the utility of ATProto is extremely narrow and will likely lose to something with a different philosophy.

That is about to change very, very soon though: https://dholms.leaflet.pub

Or you just use a different protocol for private stuff. Square peg, square hole. Round peg, round hole.

> The ATProto was designed around all data being public.

I might turn this around to say that ATProto was the square peg trying to fit into the round hole (the vast majority of people want privacy)

The main need is strong authentication, certainly for enabling P2P microblogging type apps.

The privacy stuff is covered by Signal etc. anyway, and you could layer that on top if you wanted to, but in a P2P setup privacy is always going to be a bit disappointing as so much metadata is inevitably exposed unless you do crazy stuff like flood fill the network with every message without regard to routing.

The definition of privacy is a spectrum, with what you describe at the most information secops end. Most people are not so worried about metadata leakage (they don't even know what it is, similar to atproto for most bluesky users).

What most want is to not be reposted and ratio'd by terminally online people like Blue MAGA did on Bluesky. They also just want to post to their followers or friends, not the world.

So in your concept if you are A and have two folowers B and C that don't know each other but both comment on your post what should happen? Should you see both but they not see each other?

This is the tip of the iceberg of boring questions that fall out, all of which have annoying user experience implications which are intimately tied to the underlying infrastructure and cryptosystems, effectively setting them in stone. There are reasons these systems fall into the camps of either federated with strong ID or really very centralized.

> what should happen?

A core principle of ATProto is that the cost of switching is low, thereby facilitating real competition on social media. So the answer is two parts

1. App views decide permission parameters for their modality and frame the user experience

2. Users use these permission parameters to dial in their personal or community experience

I believe there is more middle ground than you leave space for. See systems like ReBAC/Zanzibar, Anonymous Credentials, and UCANN. Google Workspace is a good model that many many people understand and are happy with (IAM wise).

With a design like this, you can build an enforceable world-read, mutual-following-write version of Bluesky, something users ask for every time Bluesky announces any new feature.

> 1. App views decide permission parameters for their modality and frame the user experience

Yeah, so this is the problem. A system flexible enough for that will be some combination of centralized, hilariously over complicated, massively inefficient, and full of odd unexpected behaviors.

Perhaps if you are after public square social media. If you are after small social media, the usage patterns and tradeoffs look different.

Modded Minecraft (smp) servers are interesting prior art for me.

the first version of bluesky was an internal effort to make twitter decentralized. atproto comes from twitter/ig type social media where almost all content is public so not supporting private data for v1 was an obvious choice.

they only decided to expand into other domains after the elon takeover made people move away from twitter and bluesky got popular way faster than expected.

> obvious choice ... decided to expand into other domains after

I'm not sure what the obvious answer is. I imagine the user insights at Twitter would be very informative, but was never on the inside to know. It's also worth noting that many members of the team joined on after ATProto was broken out from Twitter. Only they can provide an accurate recount of the information they had at hand and the decision making process around it.

Reply on news.ycombinator.com