back

by danabramov·3mo ago·view on hn ↗
Atproto isn’t “many servers sending messages to each other”. It’s structured more like RSS:

1) there’s an app-agnostic hosting layer (and anyone can run a host, a bit like personal site with RSS)

2) then there’s apps, which aggregate over data from all hosts (a bit like Google Reader or Feedly)

So there’s no such thing as “defederating”. You don’t have many copies of Tangled beefing with each other. It’s more like you can run your own hosting for your own data (if you want), and anyone can build an app that aggregates from everyone’s data (Tangled is one such app).

If this got you curious, I have two longreads: https://overreacted.io/open-social/ (conceptual) and https://overreacted.io/a-social-filesystem/ (diving into the data model).

3 comments
> Atproto isn’t “many servers sending messages to each other”. It’s structured more like RSS

Except that, crucially, RSS/Atom plays well with static nodes (e.g. personal websites generated with Jekyll/Hugo/whatever—or even written by hand[1]), and Atproto does not. (Nor does Mastodon; previously: <https://news.ycombinator.com/item?id=30862612>.)

It'd be great if the complexities needed to support the "Atmosphere" were widely recognized/acknowledged to be overkill and soon enough ended up going the way of things like CORBA and WSDL while in its place a resurgence of interest in the Atomsphere emerged.

1. <https://m15o.ichi.city/site/writing-atom-feed-manually.html>

Atom is pull, Atproto is push.

Atom was designed for news, before social media existed, where 15+ minute polling times were (borderline) acceptable. Atproto was designed for social media, in an age of Twitter users getting their news in seconds, to the point of being able to comment on live events play-by-play. There's no coming back from that world.

With that said, I wish both Mastodon and Atproto supported opt-in pull-based, static sources.

> Atproto was designed for social media, in an age of Twitter users getting their news in seconds, to the point of being able to comment on live events play-by-play.

And this is widely recognized by now to have been a very bad thing, even/especially those most susceptible to its draw. It's strange that you're framing it as a strength and not a lament.

> There's no coming back from that world.

You can't say that when everyone just begs the question and shoves application-server-needed-here protocol designs to the fore.

> And this is widely recognized by now to have been a very bad thing

It has upsides and downsides. The ability to live-post an event, or get up-to-the-minute news, can be a good thing.

Because this is about synchronizing code and not realtime social media? Social media is just an example application of the technology.
> I wish both Mastodon and Atproto supported opt-in pull-based, static sources.

Bridgy web (brid.gy) does that somewhat. Converting static sites into site.standard atproto records and whatever the corresponding activitypub format is.

There's always some Gemini protocol faction that shows up to yell that everything is wrong and we have to keep hand assembling our packets by hand or it'll never work.

Atproto's PDS is the root idea that everything extends off of, is the "social filesystem" that you control. There's a protocol objective to be able to spread your data around widely and for folks to be able to cryptographically check that that data came from you (even if you have to change hosts or even if someone sneakernets your data around). That's going to have some complexity! But it allows aggregation, is essential to how we are able to syndicate data so widely in atproto. It's so important it's in the name: Authenticated Transfer protocol.

And that in turn enables systems like Tangled here to be built, that layer stop the personal data servers, and relays. These work because there is identity.

If you need your static site to be on atproto (yay!), you can just have one of the various PDS hosts (such as Bluesky or eurosky or black sky or npmx) host the PDS for your. Since it is authenticated and user sovereign, you can permissionlessly move to a different host whenever you please, should that go awry. It's unclear to me why static site needs are an interesting or useful target that social networking ought conform to.

If you want to make a simpler network where we don't have those guarantees, please go right ahead. It feels to me like a snap reaction though that doesn't bother weighing what we have gotten or why things are this way, that is reflexively demanding.

> If you need your static site to be on atproto (yay!), you can just have one of the various PDS hosts (such as Bluesky or eurosky or black sky or npmx) host the PDS for your. Since it is authenticated and user sovereign, you can permissionlessly move to a different host whenever you please, should that go awry.

These seems to defeat the purpose of the relative amount of sovereignty that hosting a static site gives you compared to depending on a PDS.

> It's unclear to me why static site needs are an interesting or useful target that social networking ought conform to.

How is this possible?

Your data is still signed by you, and you still have the keys to move your PDS no matter what happens to your host. Do you have an actual threat model or reason why you are so afraid / unwilling to accept any compromise?

Your lack of a reply at the end, refusing to support basically your entire ask with even a modicum of supporting cause, feels a bit vindicating, that indeed you are a hostile agent & not here to engage or discuss, but to throw bombs.

> Do you have an actual threat model or reason why you are so afraid / unwilling to accept any compromise?

Have we met?

I host my PDS on a server from Racknerd that costs me $15/year
The web is already structured like this. You can poll a URL for updates. You can host your own data. Anyone can build an app that aggregates from everyone's data.
Yes, all of those things are possible. Now imagine a protocol built from the ground up for those purposes, not just possible, but the entire community and ecosystem embracing those things.
you mean http?
We've tried that, multiple times. Semantic Web, "everyone has an API" and more before and after, none of them gain sufficient traction to stick around and be built on top of.
> Anyone can build an app that aggregates from everyone's data.

Unless they gut or disable their API, and then ban scrapers, although I understand why after seeing bots crush sites.

Who are “they” here? Hosting providers? Hosting is app-agnostic and you can run it yourself.
Except it is intentionally unstructured. AT proto is one way to add the structures we obviously care about for "social" internet media on top of the freedom that the web as a whole provides.
Thanks, that does seem better for this use case!