Gregory's Server@rauschma h'mmm...
"Federated vs. centralized: Defining this difference for social networks is tricky because there are always some centralized aspects"
What do you consider to be the "centralised" aspects of #Fediverse/#ActivityStreams?
|
15 comments
 @rauschma domain names are centralised only to the extent that the domain name system is centralised – which is to say, hardly at all. There are only a few well known root name servers, it's true – but actually the system could do without them; and the rest of the system is entirely decentralised.  @simon_brooke Again, I agree! Improvements I’d love to see: Both would help with moving between servers. #1 seems easy. Not sure how difficult #2 would be. @rauschma also: "I’ll contrast Mastodon’s architecture with Bluesky’s (another social network with a different protocol that also supports decentralization)." To what extent does AT protocol actually support decentralisation? The relays are eye-wateringly heavyweight – there's no way they can be financed and run by volunteers, and there's no mechanism through which a commercial third party could monetise them. @rauschma also (I know I'm being picky here – I'm not being hostile, I'm also interested in ways we can improve social media architecture) "We need search engines for the Fediverse. Those can provide global search and algorithms for global discovery" Do we? Why? This seems directly counter to GDPR principles – people shouldn't be forced to be more 'discoverable' than they choose to be. Search engines have a lot of power and are very useful – for harm (e.g. stalking, fraud) as well as good. @simon_brooke Search indeed has to be opt-in (as it already is for server-local search). I’m fine without algorithms and better discovery and it’s awesome that the Fediverse doesn’t need a firehose (=global view, like Bluesky needs). But some people find it difficult to find interesting content and accounts. Fediverse search engines can help them. @rauschma some sort of cryptographically signed key ought to be possible. It would probably be very difficult to devise a scheme of reasonably compact keys that would guarantee that two different users could ever generate the same key, but you could make is statistically very unlikely. @rauschma Given that choice, tying user ids to server ids, especially given that we now have pretty good user migration tools, looks like a very reasonable choice. #2/2 @simon_brooke One simple approach we could copy from Bluesky: @rauschma That's an awful lot of complexity which gets you back to the same dilemma you had in the first place: either the user owns their own domain name (hosted on the domain name system) *and* a server they can point that domain name at; or they're dependent on someone else to host it. I don't think you win anything at all. @simon_brooke @rauschma If the registry with which you have registered your domain disappears, (for technical, legal, financial etc. reasons), then you have absolutely lost your identity. This is the same problem. Domain registries are these days just commercial entities like any other, and they're frankly often not at the more ethical end of the commercial entity spectrum.  @simon_brooke @rauschma |
@simon_brooke For example: domain names as used by servers. Some aspects are pluggable (arguably less centralized) in Bluesky which, at its core, is relatively centralized.