3 ms·

IMO those problems are better solved by SOLID https://solidproject.org/ https://solidproject.org/ pushed by Tim Berners Lee himself. As someone who worked(s) a

IMO those problems are better solved by SOLID https://solidproject.org/ https://solidproject.org/ pushed by Tim Berners Lee himself. As someone who worked(s) a lot on diaspora, I can tell you decentralization is hard. I don't see how atproto solve most of the complicated problems (discovery, scalability, etc.)

jauntywundrkind 1mo ago
> I don't see how atproto solve most of the complicated problems (discovery, scalability, etc.) Oh! So you don't know very much! Well, happy to help give you some links so you can learn some of the basics! @pfrazee gave a talk recently on how atproto was built from day 1 for scale. It's not a secret, it's in the name, Authenticate Transfer Protocol! Your key that you control signs all of your content! Such that your can be forwarded by anyone, that it continues to bear your authentication, wherever it goes, whatever replicas it's on! https://youtu.be/BoJnj2yPf14?is=YcnoJZSqQw1YBhl9 https://youtu.be/BoJnj2yPf14?is=YcnoJZSqQw1YBhl9 It's so easy and small that one can run a whole network firebose off a small VPS! There's a list of them. https://atproto.at/relays https://atproto.at/relays Discoverability is a function of this! There's all sorts of tools to filter for different kinds of data. There's indexes, so you can find things like back links, https://www.microcosm.blue/ https://www.microcosm.blue/. And there's specific schemas (lexicons) for helping find blog posts, https://standard.site https://standard.site. I'm a fan of the API that Solid exposes, I'm a fan of the json-ld. But for discoverability and scalability, in terms of being a networked protocol, ATproto is great.
- flaburgan 1mo ago
  I would need a real chat to discuss this in depth and certainly can't do this here, but this is not what I meant by scale and discoverability. Scale is, how can I run my own server with access to all the content without having a DB that explodes, and discoverability is the usual "if you don't have the full ID of someone you can't find them". The attempts to solve these are old as the internet and every solutions that I know of (such as relays) means re-adding some centralization in the architecture. I still have to see a fully decentralized protocol which is also reliable and fast (so, it scales, even for small servers). Matrix is the closest to that but they are facing tons of problems which come from the architecture itself.
  - jauntywundrkind 1mo ago
    You can literally mirror all Bluesky on a medium/small-ish server, with Hubble. The design scales. There's a number of different tools to find accounts out there. The main way apps do account type-ahead is a free 3rd party service, one you could run yourself. No one does non-centralized because the network is easy to index & there's no use case for trying to ask everyone each time to do the work, that's a bad strategy, why do it? Kinda feel like you are just blowing chaff, moving goalposts, not interested in finding out or learning. So, yeah, seems like we're done.
    - cindyllm 1mo ago
      [dead]

jauntywundrkind 1mo ago

> I don't see how atproto solve most of the complicated problems (discovery, scalability, etc.) Oh! So you don't know very much! Well, happy to help give you some links so you can learn some of the basics! @pfrazee gave a talk recently on how atproto was built from day 1 for scale. It's not a secret, it's in the name, Authenticate Transfer Protocol! Your key that you control signs all of your content! Such that your can be forwarded by anyone, that it continues to bear your authentication, wherever it goes, whatever replicas it's on! https://youtu.be/BoJnj2yPf14?is=YcnoJZSqQw1YBhl9 https://youtu.be/BoJnj2yPf14?is=YcnoJZSqQw1YBhl9 It's so easy and small that one can run a whole network firebose off a small VPS! There's a list of them. https://atproto.at/relays https://atproto.at/relays Discoverability is a function of this! There's all sorts of tools to filter for different kinds of data. There's indexes, so you can find things like back links, https://www.microcosm.blue/ https://www.microcosm.blue/. And there's specific schemas (lexicons) for helping find blog posts, https://standard.site https://standard.site. I'm a fan of the API that Solid exposes, I'm a fan of the json-ld. But for discoverability and scalability, in terms of being a networked protocol, ATproto is great.

flaburgan 1mo ago
I would need a real chat to discuss this in depth and certainly can't do this here, but this is not what I meant by scale and discoverability. Scale is, how can I run my own server with access to all the content without having a DB that explodes, and discoverability is the usual "if you don't have the full ID of someone you can't find them". The attempts to solve these are old as the internet and every solutions that I know of (such as relays) means re-adding some centralization in the architecture. I still have to see a fully decentralized protocol which is also reliable and fast (so, it scales, even for small servers). Matrix is the closest to that but they are facing tons of problems which come from the architecture itself.
- jauntywundrkind 1mo ago
  You can literally mirror all Bluesky on a medium/small-ish server, with Hubble. The design scales. There's a number of different tools to find accounts out there. The main way apps do account type-ahead is a free 3rd party service, one you could run yourself. No one does non-centralized because the network is easy to index & there's no use case for trying to ask everyone each time to do the work, that's a bad strategy, why do it? Kinda feel like you are just blowing chaff, moving goalposts, not interested in finding out or learning. So, yeah, seems like we're done.
  - cindyllm 1mo ago
    [dead]