@Natureshadow@floss.social
It's tricky because ActivityPub made design choices that put instances, not users, at the center of its operation.
It's really hard to fit distributed identities into such a system.
@h@mymath.rocks @pinkyfloyd @smallcircles @lrhodes @bengo @liaizon
There are downsides to that, though, particularly when it comes to scalability, under the ActivityPub design.
Some aspects of AP scale by instance more than user, so going with one user per instance can really increase resource costs of operating the system.
@Natureshadow@floss.social @bengo @lrhodes @h@mymath.rocks @smallcircles @liaizon
absolutely but that's because the original software was written with large instances in mind.
i'm dumb and i spun up a pleroma instance on Digital Ocean on my own that costs me $12 a month and i'm having a spectacular time with my 25GB storage space.
which to me implies that with a little tweaking, in bulk this service probably costs about $0.20 per user per month.
https://cloud.payfrit.com/index.php/s/65wQKWWe5CJPqcH
No, it's not an implementation issue but the design of the protocol itself, the way ActivityPub has instances address each other and send out per-instance messages.
The protocol allows bundling of messages to multiple users on one instance, but with one user per instance the messages would need to be duplicated, scaling with instance count.
@Natureshadow@floss.social @bengo @lrhodes @h@mymath.rocks @smallcircles @liaizon
you might have missed where i mentioned "a little tweaking" :)
i know that google gives 25Gb free (plus some CPU activity, etc) to any bonehead with an mail address.
i know that the current ap model has gotten us to the point where instances simply just die because admins can't afford the money or time costs. that's not a feature.
software evolves.
This isn't something that can be solved with a little tweaking, though.
As it is, we'd expect every post to be sent to every subscribed end user, by sending a copy to each relevant instance with a list of recipients. That's the core heart of the protocol, no matter what language software may be written in, how it organizes variables to save memory or whatever.
But that requires a load that increases as # of posts * instances.
It ends up looking awfully exponential, a graph that quickly explodes with increasing use.
It's no simple feat to tweak the system to arrest such a problem of exponential scaling, but switching to single user instances only drives the explosion faster.
@Natureshadow@floss.social @bengo @lrhodes @h@mymath.rocks @smallcircles @liaizon
i understand all of that.
i know that all of the tools are already written in order to make what i have described happen.
i personally don't believe AP has a very robust future.
https://cloud.payfrit.com/index.php/s/65wQKWWe5CJPqcH
exactly and i for one believe that the default case should be every user having their own instance.