/fit/ thicc fairy 🤬 is a user on sunshinegardens.org. You can follow them or interact with them if you have an account anywhere in the fediverse. If you don't, you can sign up here.

Writing a paper titled "ActivityPub: from decentralized to distributed social networks" for the Rebooting Web of Trust summit.

If you wanted to look at how AP could be used for E2E, how to remove SSL and DNS centralization from the system, etc you might like this paper.

I updated the paper with typo fixes, clarification on vocabulary of "decentralized"/"distributed" (linking in a diagram), and some comment on incremental improvements as relief from "great is the enemy of good". gitlab.com/dustyweb/talks/blob gitlab.com/dustyweb/talks/raw/

@cwebber Haven't read it all through yet, I'm all for the hot takes.

> Consider this a lesson that a protocol alone cannot build a distributed network; the community must build and maintain a healthy number of nodes and avoid the temptation to let a few large providers control the space of a federated network.

I think this is one of those social problems that have a technical solution, or rather, it's not a purely social problem. As Dominic Tarr said the other day:

> sure, the design of email is decentralized (federated) but the usage is centralized (or centralized-ish) I'll give you that. How do we design a system that does not tend towards centralization? I guess there needs to be something pulling the peers apart, instead of together?


Having ideals to avoid temptation isn't a long-term solution. The built-in forces of the technology must push peers to the edges rather than to the center. Community ideals do not scale up.
@cwebber On the other side of the coin we have https://identi.ca/mlinksva who has said a couple of times that large flagship nodes are the only thing that could be powerful and convenient enough to fight the big social silos. I'm not so sure.

Regardless if we try to solve centralization with federation or we believe true peer-to-peer is key, we need to solve #userops. From what I read, the people who are closing their OStatus nodes these days, be they Mastodon or GNU Social, are mainly doing it due to the burden of maintaining them.

@clacke I've heard Mike Linksvayer say that and there's hardly anyone whose opinion I respect more than mlinksva :) But, I'm not sold. Gitorious is a good example of where I get nervous. But it's true that the C2S model does make "hosting a lot of small nodes" not easy or likely to survive either.

The Right Thing to me seems to be, in the long run, two things:
- make deployment and long-term maintenance considerably easier for every day users (hence the UserOps stuff)
- move incrementally from a c2s model to a p2p model using content addressed storage, etc. I tried to show in that paper how we could incrementally move there without throwing out what we have and what works

@cwebber @clacke so you're saying people should use :DDDDDDD

@xj9 Of course!

What is grid? Not exactly the googlebliest term.
/fit/ thicc fairy 🤬 @xj9

@clacke 🤔 I probably need to consider SEO when I'm naming these things tbh

is a scheduler along the lines of fleet or k8s the idea is to turn your personal devices (+ optionally cloud machines) into a single distributed UNIX-ish thing uses for isolation and packaging not ready to share yet, but it *is* functional enough that I can use it to host my personal cloud nonsense


@xj9 You don't seem to know exactly what it means yet, but I like the words that come out of your keyboard.