@soapdog Yes, lots of details to work out for our use cases, but the pull model is what makes RSS so crazy powerful and scalable. Push models are resource hogs that approach exponential growth in a large network like the fediverse — the medium-sized site that hosts my account is constantly struggling to stay one step ahead of the fedi's surging resource demands.

@david_megginson that is the exact problem that I want to get rid of. It really sucks. For example, I'd like to host a single-user AP server here at my flat, but knowing that it will use so much bandwidth is enough for me to reconsider it.

@soapdog Why would it make it easier to self-host? Dealing with pull requests is on the whole more difficult than a simple send-and-forget. Tracking who is in on a conversation is easy enough. On the other hand, tracking every conversation you've ever been in is way more difficult, and wasteful pulling from every single one of them. It wouldn't work.

@soapdog Part of the problem is that we set the expectation that the fedi would work just like a monolithic social media site. If we reduced that expectation just a little, the demands would go down a lot.

Users already accept (or fail to notice) that Mastodon's search and hashtag features aren't really global across the fedi.

@soapdog There's a poll-based version specced at https://fediverse.codeberg.page/fep/fep/b06c/, sadly with no notable implementations (wouldn't be interactable by Mastodon etc.), but it's an opportunity to break new ground as an implementer if you know anyone who'd like to experiment with it.

#ActivityPub #FediDev

@julian that was a very interesting read. I think I can implement this. Will try a small experiement soon. Thanks for sharing.

@david_megginson the good thing about expectations is that we can adjust them

I'm not saying replace AP with something else, I'm saying let's build something different and let them all blossom. I really liked the spec described here:

FEP-b06c: ActivityPoll - Fediverse Enhancement Proposals

ActivityPoll is a proper subset of ActivityPub that excludes activity delivery, making it easier to implement for static Web sites or content management systems. It meets an equivalent need to RSS or Atom feeds.

(fediverse.codeberg.page)

@soapdog If you do, be sure to tag @evan! As the FEP author I imagine it would be exciting to hear about new implementations.

@soapdog

We're using pull semantics for data exchange in the federation of wikis, if you'd like to try that out. #FedWiki

via @smallcircles

@julian @soapdog ofc, although I think the overemphasis on number of implementations for FEPs is a bad idea. Most standards docs, like RFCs and W3C recs, don't include implementation info. It changes over time!

@evan @soapdog True. I was talking about this just yesterday, if we wouldn't be better served by something wiki-like to host a software project × FEP implementation matrix instead of asking FEP authors to keep their lists updated.

That aside, in this context I'm thinking about how more implementations in the wild can motivate other implementers. My own website still isn't AP-capable (it's planned, but low priority). If I knew there were a sizable ActivityPoll user base, I'd more likely use it.

@julian @soapdog I agree, it really lowers the bar for publishers.

@soapdog makes sense. I never had fear of not being able to selfhost an RSS feed

@soapdog@toot.cafe hmm... just thinking aloud here.

You posit in another post that the network effects inflate exponentially:

> Push models are resource hogs that approach exponential growth in a large network like the fediverse

That's not true. If you post a message then it sends a copy to each follower. That's linear growth. If you collapse recipients via shared inboxes you can reduce that further.

If you're referring to the torrent of requests that happen if your post is shared (the "thundering herd" problem) then that's actually a PULL happening from those requesting instances!

Secondly, in a pull model of AP, you would need to continually poll servers of all your followers so as to approach a real-time effect. You'd be polling servers over and over again, and many of them would have nothing new, with so much wasted traffic.

If your expectations include semi real-time updates, the push model is much more performant, in my humble opinion.