For example, for me, here are some things I wish to see (or would implement in my design) :
- design around ease of self-hosting. A non technical user must be able to self host easily and at a very low cost.
- Embrace content sorting and filtering algorithms, but on the client side, with optional control by the user.
- Standardize tags on all content. So many of the different ways different platforms classify or organize content can be implemented as tags, which increases interoperability between them.
- Abandon obsession with real-time-first implementations for use cases that don’t explicitly need it.
- Transferable user identity (between instances)
- User identity and authentication as separate service from social network instance
Would love to hear yours!
You can only filter and sort what was downloaded by the client. So that runs into resource constraints.
I’m so with you. https://xkcd.com/927/
That’s more the ATproto/Bluesky vision.
In the client, you wouldn’t need to be sorting and running extensive calculations on all data. You could, e.g, build the front-page by indexing/scoring posts and comments that have been created since your last visit with a hard cap on some time window (last 48h) or total data points (e.g, keep only the most recent 10k objects in a local hot database, freeze/archive everyhing else.)
That’s what RDF/JSON-LD gives us for free. There is no need to have us arguing over what each tag means, all that developers need to do is to learn how to use the different vocabularies.
Doesn’t mean that we can’t adopt it.
Absolutely. There’s a lot you can do. The “For You” Feed on Bluesky is quite instructive. https://bsky.app/profile/spacecowboy17.bsky.social/post/3mb2r5qei322a
But when you’re talking about sending a lot more data to clients, you really need to consider what that means for the internet bill of instance owners.
I would argue the opposite, actually. A lot of this data could be distributed in a p2p manner and the client nodes would have to rely even less on the servers. The key part would be that this data would have to be self-authenticating, but we do have the mechanisms to do that (Linked Data Signatures)
Good pitch. You could also ask people to help out with the more expensive computations. Say, adding alt text.
There shouldn’t be resource constraint issues in downloading new content Metadata from all your subscriptions
When I say “standardize tags on all content”, what I mean is make it a standard to have the option of having tags on every type of content, and treat tags as a first class attribute of the content. The XKCD you reference is not relevant.
Sure, they use it. But it is compatible with activityPub as well.