I'm coming to the conclusion that community-owned and operated small clouds (co-ops) with easy onramps for self-hosting open source services like mail, storage, and VPN are the only way forward.
-
@mttaggart @HolosSocial i disagree with making it 100% turbo decentralized and exist kind of like a cloud in the middle of nowhere because that has been tried in stuff like matrix and others and it just has led to a lot of different issues, mostly because of how media proxies work and how it entirely destroys any sense of moderation.
i still think it is a good idea to create something more akin to... towns, that specific people with enough time can dedicate their resources to mantaining, so that other people can join freely and adhere to existing rules, kind of like the fediverse, but removing the "black box" element of "you just join and it just works" even though it's a federated system
i think the way through is not unification, but intentional separation of the presentation and rules and structure of each instance, but total unification of the protocol and basics so that people can choose their own client and technology can be simplified to them
i think that... there's also a nomenclature issue in how we manage this kind of thing, right? where we call it "the fediverse" and "federated" and "decentralized" on "servers" that "speak activitypub" and use the "mastodon api" for the "client", it's just very confusing when you frame it as open source tech specifications when it should be treated instead with analogies and metaphors, which is how the early days of personal widespread computing got their way with letting people understand how the underlying tech works: the wallpaper, the files, the desktop, the recycle bin... all of these are analogs for real life objects
@nelson@wetdry.world @mttaggart@infosec.exchange @HolosSocial@mastodon.social i think the way atproto does this is done very nicely, its very easy for non technical users
now it is rather decentralized right now as most people aren't too aware of the fact they can run their own pds. but thats not really a fault of the protocol -
I'm coming to the conclusion that community-owned and operated small clouds (co-ops) with easy onramps for self-hosting open source services like mail, storage, and VPN are the only way forward. Every corpo service is eventually going to make you ashamed to use it.
@mttaggart @zackwhittaker
I'm spending my efforts in this direction, too. -
Not for nothing but I've written a very well-regarded guide on home labs if you want to get started.
@mttaggart how does one access the book to download, it again when they purchased it a year ago or so and the download link in the email is not working anymore?
-
@mttaggart @brahms my personal opinion is biased, but I like old CAD & developer tower workstations. You can frequently get them with Xeon server processors (and I’m sure AMD equivalents, but I last bought nearly a decade ago), and the cooling is set up to be effective without being a menace on/below someone’s desk, plus you can fit a respectable number of 3.5” drives in them. OTOH, if you wanna play with clusters, I’d look at micro desktops or laptops (built in UPS), cooling them can be more situational, and you’re probably limited to fewer & lower capacity internal drives, or using external USB drives.
@ajn142 @mttaggart @brahms I've bought two M720q (16GB, 256GB M.2 + 256 SATA each) as a base for HA proxmox cluster. Power consumption is minimal and it even runs some small LLM for weather-json-to-human-text. Those were probably under 100€ each.
-
@ajn142 @mttaggart @brahms I've bought two M720q (16GB, 256GB M.2 + 256 SATA each) as a base for HA proxmox cluster. Power consumption is minimal and it even runs some small LLM for weather-json-to-human-text. Those were probably under 100€ each.
(and yes, that's a shoe rack with thin plywood as a shelf
) -
I'm coming to the conclusion that community-owned and operated small clouds (co-ops) with easy onramps for self-hosting open source services like mail, storage, and VPN are the only way forward. Every corpo service is eventually going to make you ashamed to use it.
@mttaggart Or be acquired by someone who'll do that. Eg. Linode being absorbed by CloudFlare.
-
I'm coming to the conclusion that community-owned and operated small clouds (co-ops) with easy onramps for self-hosting open source services like mail, storage, and VPN are the only way forward. Every corpo service is eventually going to make you ashamed to use it.
Interesting approach. Are there concepts available for a secure, operable, reisilient decentrralized mail system? Which can be installed an run without having a master degree in IT? Some kind of fedi-mail?
-
@mttaggart I think email especially, without that level of resiliency, is basically malpractice
@delta_vee @mttaggart
do large-scale email servers (gmail, etc) still retry after failed delivery like the smaller ones do? -
@falk_ That's true, but indefinite support may be an unreasonable expectation. A reasonable expectation may be an exit agreement.
@mttaggart @falk_
unfortunately, "indefinite" is the only reasonable expectation in terms of email addresses
you could have everyone bring their own domain tho
-
I'm coming to the conclusion that community-owned and operated small clouds (co-ops) with easy onramps for self-hosting open source services like mail, storage, and VPN are the only way forward. Every corpo service is eventually going to make you ashamed to use it.
@mttaggart @zackwhittaker Maybe not “every” corpo service… https://www.infomaniak.com/en/sovereign-cloud
-
@mttaggart I have heard of this model being used in Barcelona. I think that @brunovianna might be able to provide further information.
@plwt @mttaggart yes, we started a community server with mail, cloud, local ai and other services in my neighborhood. But it's not easy to get people to use it

-
@delta_vee @mttaggart
do large-scale email servers (gmail, etc) still retry after failed delivery like the smaller ones do?@wolf480pl @mttaggart Sometimes, but retrying an hour later doesn't help the user get their login code
Modern email is, probably unfortunately, too deeply enmeshed in modern access flows to really permit the kind of eventual reliability it was originally built around
-
@plwt @mttaggart yes, we started a community server with mail, cloud, local ai and other services in my neighborhood. But it's not easy to get people to use it

@plwt @mttaggart but I think we should find other types of governance not necessarily attached to the territory like us. I made a proposal to discuss it in Mozilla festival and it got accepted
-
I'm coming to the conclusion that community-owned and operated small clouds (co-ops) with easy onramps for self-hosting open source services like mail, storage, and VPN are the only way forward. Every corpo service is eventually going to make you ashamed to use it.
@mttaggart I am also one of those running servers for friends (mail, web, Nextcloud, Jitsi Meet, SearXNG etc) and I think about moving this to a private community, so I am not the only one taking care of the infrastructure.
-
@ireneista @mttaggart @vfrmedia - so I think my take on this is more community first, tech second. Like, if the output is basically a support network and a set of tools to help individuals and community organisations self host, then I think I'd be quite happy with that.
Or maybe the solution is lower level services, e.g. shared server management with tools to make it easy for individual users to spin up their own Mastodon instance on their own domain.
VPNs are possibly a bit of a special case, and maybe the DEFCON folks reached the right conclusion there when they decided just to support Tor.
> "support network and a set of tools to help individuals and community organisations self host"
we have yet to articulate this with @coopcloud but i do believe we are going in this direction
the solidarity network is crucial to emphasise which goes beyond hollow top-down claims of "reuse" and "community"
unity upon strategic tool use has major benefits which stands in stark contract to the dominant reinvent the wheel tech hype cycles...
-
@nelson@wetdry.world @mttaggart@infosec.exchange @HolosSocial@mastodon.social i think the way atproto does this is done very nicely, its very easy for non technical users
now it is rather decentralized right now as most people aren't too aware of the fact they can run their own pds. but thats not really a fault of the protocol@stag @nelson @HolosSocial So I think the PDS model of data transferability is neat. Blacksky shows what it takes to do all this without reliance on Bluesky. It's far from simple.
-
@stag @nelson @HolosSocial So I think the PDS model of data transferability is neat. Blacksky shows what it takes to do all this without reliance on Bluesky. It's far from simple.
@mttaggart @stag @HolosSocial there are actual new attempts of using atproto without any relays and external architecture, things like "red dwarf" and even "wafrn" now adopt this mechanism
i am not exactly sure on how it works but as far as i understand, it performs activitypub-style federation but with proper backfilling and it's pretty darn efficient for small and big deployments alike
maybe my friend @alexia is better at explaining it than i am, she does work pretty closely to wafrn
-
@mttaggart @stag @HolosSocial there are actual new attempts of using atproto without any relays and external architecture, things like "red dwarf" and even "wafrn" now adopt this mechanism
i am not exactly sure on how it works but as far as i understand, it performs activitypub-style federation but with proper backfilling and it's pretty darn efficient for small and big deployments alike
maybe my friend @alexia is better at explaining it than i am, she does work pretty closely to wafrn
@stag@mk.absturztau.be @HolosSocial@mastodon.social @mttaggart@infosec.exchange @nelson@wetdry.world
hi hello
to TL;DR this because I don't want to make too large of an explanation, for a good few months now instead of relying on a relay+appview we've instead been relying on https://www.microcosm.blue/ and more specifically Constellation as well as Jetstream (both of which amount to basically filtered down relays) which dramatically decreaess the amount of storage use, network bandwidth and CPU required to run wafrn, even if you were to do it full-stack setup with hosting constellation, jetstream, wafrn and a PDS on your own.
I think jetstream still uses a relay as an upstream but even those have become one hell of a lot cheaper to run thanks to more customizability to what is and isn't kept and for how long
point is all of the above can now run on a really shitty contabo vps for maybe like 3-4€ or whatever @gabboman pays
(that said, depending on circumstances it can still be more expensive than an activitypub server)
(do also note that the most expensive part of Bluesky's stack has always been the AppView, which provides an API, caching with redis, full-text indexing, a CDN…) -
@stag@mk.absturztau.be @HolosSocial@mastodon.social @mttaggart@infosec.exchange @nelson@wetdry.world
hi hello
to TL;DR this because I don't want to make too large of an explanation, for a good few months now instead of relying on a relay+appview we've instead been relying on https://www.microcosm.blue/ and more specifically Constellation as well as Jetstream (both of which amount to basically filtered down relays) which dramatically decreaess the amount of storage use, network bandwidth and CPU required to run wafrn, even if you were to do it full-stack setup with hosting constellation, jetstream, wafrn and a PDS on your own.
I think jetstream still uses a relay as an upstream but even those have become one hell of a lot cheaper to run thanks to more customizability to what is and isn't kept and for how long
point is all of the above can now run on a really shitty contabo vps for maybe like 3-4€ or whatever @gabboman pays
(that said, depending on circumstances it can still be more expensive than an activitypub server)
(do also note that the most expensive part of Bluesky's stack has always been the AppView, which provides an API, caching with redis, full-text indexing, a CDN…)@alexia @stag @gabboman @HolosSocial @nelson
Right. Jetstream does rely on an upstream Relay, and I think Bsky acquired Jetstream from an independent project. And yes, the AppView for large lexicons is the hard problem.
-
@alexia @stag @gabboman @HolosSocial @nelson
Right. Jetstream does rely on an upstream Relay, and I think Bsky acquired Jetstream from an independent project. And yes, the AppView for large lexicons is the hard problem.
@nelson@wetdry.world @alexia@app.wafrn.net @mttaggart@infosec.exchange @HolosSocial@mastodon.social @stag@mk.absturztau.be
bsky centralized discourse again?
gona be quick.
each wafrn instance in https://join.wafrn.net is independent of bluesky.You can host a wafrn instance on a shitty contabo vps or even worse (gabboman xyz costs less than 20 euros a year).
We use external apis that are easily hosteable on a VERY EXPENSIVE raspberry pi. such extreme compute power required.
its a different architecture. and I will say what I always say: knowing how one works makes harder for your brain understanding the other one.
I say this as someone who has wrote implementations of both