I'm sorry, it took *how* many servers to post a single long message from Ghost to 5k fediverse accounts and handle some replies?
-
Jenniferplusplusreplied to Jenniferplusplus on last edited by
@polotek @hrefna @thisismissem @kissane @fediversereport @hongminhee
There's many very reasonable optimizations that could be made, and would be conformant with the spec. But that really just means it's not prohibited, not that it's supported. You could batch messages, for instance. That's a valid thing to do in AP docs. But I would have no faith that any other software would know what that means or even be able to deserialize it in a useful way.So, no batching.
It's the same in so many ways.
-
@bengo @johnonolan @kissane @fediversereport this makes a fair amount of sense to me, shame it was removed
-
@jenniferplusplus @polotek @hrefna @thisismissem @kissane @fediversereport @hongminhee even within the bounds of the spec you can forsee e.g. a group of instances sharing one sharedInbox (which handles fanout between them). You could define some kind of even more optimised inbox property, if you wished.
But if this were a major problem (as opposed to being a problem of building your application in a scalable manner + tuning things like your http2 connection keep alives) I suspect that Mastodon gGmbH would have done it to keep mastodon.social working
-
Marco Rogersreplied to Jenniferplusplus on last edited by
@jenniferplusplus @hrefna @thisismissem @kissane @fediversereport @hongminhee you know, I think I need to be trying to narrow my engagement to talking about single user instances. It's becoming clear that it's a pretty different conversation than scaling large shared instances. I worry that I end up talking past people because they're responding to different contexts.
-
@jenniferplusplus @fediversereport @hongminhee @hrefna @kissane @polotek @thisismissem I think its important to understand that even at an individual level social media is actually a firehose of data much larger than most developer’s initial assumptions. You very quickly pass the point where naïve implementations of things fall apart.
Even with the protocol as it is today, implementations like Akkoma and GoToSocial scale very well to delivering posts to thousands of users (Pleroma & derivatives may be quite famous for scaling problems, but mostly those are on the side of the Mastodon API implementation; the federation queue itself screams) in quite low amounts of resources.
Now, there are other ways of implementing things, but they mostly just move the hard problem around (AP generally places the costs on the poster; ATProto puts the costs on the reader, who must follow approximately the entire firehose)
-
@erincandescent @hongminhee @hrefna @thisismissem @kissane @fediversereport @polotek
mastodon dot social essentially *is* that big shared inbox. It has like 20% of the active fediverse population. -
@jenniferplusplus @hongminhee @hrefna @thisismissem @kissane @fediversereport @polotek right, but they feel the challenge of fanning out posts to followers as much as anyone
-
@erincandescent @jenniferplusplus @polotek @hrefna @kissane @fediversereport @hongminhee there is a new multibox FEP too: https://codeberg.org/fediverse/fep/src/branch/main/fep/0499/fep-0499.md
-
@thisismissem @jenniferplusplus @polotek @hrefna @kissane @fediversereport @hongminhee This is what I’ve sometimes referred to as “enveloped inbox” by analogy to SMTP (though the primary motivation was user collections & Bto/Bcc)
-
@jenniferplusplus @fediversereport @hongminhee @hrefna @kissane @polotek @thisismissem anyway most of what i’m trying to get across is that these problems are not inevitable. Fanning things out to my 2300 followers from my single user instance doesn’t really do anything major on the CPU usage graph and I’m fairly confident it could go much higher before I started noticing anything each time I posted.
-
@kissane @fediversereport @thisismissem @bengo @evanprodromou absolutely - we’ll try to keep publicly documenting gotchas and things we struggle with, to create more shared resource and knowledge
Hopefully along the way we’ll also document some solutions
-
The Nexus of Privacyreplied to Marco Rogers on last edited by
you're not being annoying, it's a great topic of discussion. @julian has made some related points. it's so valuable getting new eyes on things that everybody had fallen into the habit of getting taken for granted. I once asked how much scaling analysis had been done in the standardization process and the answer was unsurprisingly "none".
@polotek @hrefna @thisismissem @jenniferplusplus @kissane @fediversereport @hongminhee
-
@[email protected] yikes. i've got self-hosted ghost, it's okay, i was looking forward to fedi integration ... but not if i have to spin up 10 servers.
on a different subject ... do you have time to talk sometime next week? i'm thinking about what next, and there's various stuff to catch up on as well, so it would be great to connect.
