The author compares this model to the web, but it also sounds something like GitHub, but with more cryptography. Users publish their own repositories, but you still need big, expensive services to index it all and help people find each other.
Currently the BlueSky search engine is pretty bad; it only goes back a few hours. If you don't post for a day, none of your posts will be found. Searching on obscure terms doesn't work very well.
Don't look at our search service as deeply correlated to the indexing model. That's not the point of the comparison, and our search service is quite subpar. We've only had time to put an ElasticSearch instance against the data, do some basic tuning, and then keep it online while we focus elsewhere.
Links worked well enough for the web, Google just came along and painted a picture of being able to find whatever you want whenever you want it.
Social graphs can offer a similar discoverability without search. You see what those you follow post, comment on, like, share, etc. Following tags just like you follow a person also works well. Before assuming we must have to have massive indexing and searching services, ask yourself whether you really need to be able to discover anything that anyone around the world posts, regardless of whether you have any sort of connection to that person, their context, their society, or their culture.
Hmmm. I'm sympathetic to this perspective, I've sketched ideas for social media that was more "friends of friends" oriented (specifically I want to have a platform based around collections of media - I create a collection, upload some pdfs and videos and markdown files etc, and start serving those via torrent. Peers that happen to place the same file in their own collections suddenly become co-seeders. I want to see "what other collections does this file appear in" and make friends that way. I feel like limewire used to work like this, you find someone with a grateful dead bootleg and wonder, what other files are they serving? I also use ebay kinda like this, find an interesting item and look at "sellers others items")
Anyway, I agree that a social graph has the potential to offer human-scale and $5-VPS-scale discoverability. The problem is probably with user expectation and network effect. How do you get people to use the thing when no one is using it yet, to bootstrap to the point where you might actually bump into someone with the same niche interest as you. Still, it could start with small communities same way telegram group chats and discord servers do. You just have to bring the people to you instead of just trying to index the firehose of global content.
I'm actually curious, what's a use case that really makes the broader system better by being able to discover content from anyone, anywhere?
The best I can come up with is a historical record. Beyond that I'm at a loss for a use case that is noticeably better with search compared to social graph-based discovery.
My point was that something like BlueSky can make content more discoverable by letting you build your own social graph and view content from those you're connected to.
Search is most useful for discovering content that you otherwise have no connection to at all. There is also the historical aspect and there really isn't a way around that, but in the original context of finding people actively talking about a specific topic it isn't a necessarily must-have feature.
Right my point is you’re basically describing a search engine and that requires constant downloads and storage for it. Any useful graph that comes close to current social media levels of discoverability is going to require lots of storage.
Otherwise you’re talking about a simple feed on someone else’s computer delivered to you and this defeats the purpose of federalization.
We must be talking about slightly different things here. Sure the data all has to be stored somewhere, but I'm not arguing for everyone to have a local copy of all of Twitter just so they can search it.
I wasn't pushing for federalization at all, and honestly its a bad fit for anything similar to social networks IMO. Federalization doesn't work well if the goal is universal reach, why federate the network graph when you still want to access all data on the network? The fediverse won't work at scale, the sheer amount of data duplication and network traffic would be insane. Even if we could somehow find a way to handle all the traffic, it sure seems like that would fly in the face of environmental efforts so many are onboard with (storage and bandwidth aren't cheap at that scale).
Currently the BlueSky search engine is pretty bad; it only goes back a few hours. If you don't post for a day, none of your posts will be found. Searching on obscure terms doesn't work very well.