First new #Disintermedia blog post for 2026!

https://disintermedia.net.nz/in-defence-of-servers/

Well ... kind of new. This is the first of 3 posts I wrote for the Bridge Seat co-op blog, which I'm reposting so I can stop sharing links to SS.

This one is about why geeks need to stop questing for the Holy Grail of pure P2P networks, and embrace a vision of servers as community-scale infrastructure, run by and for the people using them.

A vision of server admin of the people, for the people, by the people.

#P2P #servers

In Defence of Servers

Why pure peer-to-peer networks aren't always better than federated ones

Disintermedia

@strypey

i dont know. this seems to entirely miss the point. p2p has secure holepunching built in. it either works or it doesnt and once it works irs as good as TLS and the many other infrastructure bits you trust.

mainstream users remember their gmail/email passwors or seedphrase - if they lose it, they are done.

both, p2p and centralized can build social recovery mechanisms if you are willing to trust third parties, but in p2p you CAN go full self custodial.

...no differene heren, BUT:

@strypey

having a self custodial keypair derived from e.g. 12 or 24 words means, you can migrate you device or have multiple devices effortlessly and things just work.

your devices, dedicated cloud services or friends can backup your data - you dont have to choose and trust a single admin - and as long as at least one peer has the data, it restores effortlessly (bittorrent style).

no costs for servers, no costs for or expiration of domains either. no censorship too.

@strypey

i just went through it. fediverse signups are so much work for admins that most instances are locked down or reject your application, this demotivating ppl from joining or just defaulting to mastodon social... which becomes a mega instance.

...and gets prone to subpoenas, or bribery or other sort of attacks to at least make deals oe buy the instance and influence the algorithm or add ads.

none of this is ever an issue in p2p.
you connect with other consent based only

@strypey

but its still cool to have "blacksmiths" around to run relay servers or backup services (eg. servers which join as peers just to backup and seed data) ...any service offered on a traditional server can also be offered by a peer which can run on a home server, or hetzner or AWS - doesnt matter.

peers can connect to it ans send requests and get responses - nobody stops anyone from building anything that also exists with traditional servers ๐Ÿคทโ€โ™€๏ธ

you just get rid of e.g. DNS issues too ๐Ÿ™‚

@serapath
> p2p has secure holepunching built in. it either works or it doesnt

Mostly it doesn't, when you really stresstest it, as we've discussed before. Remember you're not talking to a noob here who just hasn't tried P2P systems. I'd wager I've tested as many as you, if not more.

The perspective in the blog post you're airily dismissing is based on hard-won experience, going back to FreeNet and SpeakFreely in the late 1990s. I used to drink the pure P2P kool-aid, now I know better.

@strypey

try keet in any network scenario.
it even punches through the chinese firewall.

for some carrier grade networks scenarios it support optional blind relay in case both peers are behind multiple layers of port randomizing routers, but it really just works.

no need to manually configure or maintain stun and turn servers.. every peer runs a dht node and its baked in.

i think you just havent tried dat.
its different from all the others.
nothing compares.

@strypey

i'm not dismissing your experience.
i have the same experience, because apart from dat, nothing is reliable.

dat is in its own league. its mature, battle tested and production ready.

@serapath
> try keet in any network scenario

I think we already talking about how this has proprietary dependencies, which makes it a non-starter.

A few days ago I asked you for DAT app recommendations, and I'm a bit behind on checking my notifications, so if you've done that, I'll check them out. If not, hit me!

Have you tried Moss, or any other messaging apps based on HoloChain? Thoughts?

@strypey

holochain architecgure is broken. they offload data into the DHT.

A dht is not meant to store large amlunts of data. mass adoption will kill this mechanism immediately. dht's should not be used for anything else than minimal and temporary information for peer discovery.

its holochains biggest downfall.
otherwise they have VCs ans a tokensale to fund their token on ethereum and is imho firmly part of the web3 scam ๐Ÿคทโ€โ™€๏ธ

...not just that, some ppl might be able to tell you more @pospi

@serapath @strypey well #Holochain has had far-right personalities leading its communities, and #Keet actively includes them, so there's that. Holochain has repeatedly championed White Nationalist hate actors and marketed itself to the far-right through its founding members and ecosystem growth manager.

IMO (as someone who collaborated with them for a long time and knows the architecture deeply) the DHT scenario and Holochain's forkable network architecture makes this a much more dangerous combination. No trust or consent is needed to replicate peer data to new group members, and data and code can be forked off to new groups any time. At least in Keet replication is up to the individual and not the volk.

@pospi @strypey

i am also not recommending keet as something to use. the dat stack is open source and keet like messengers can be built with very little effort - not so quickly to the same UX as keet, but i work woth students who have built an open source basic messenger with a ux more similar to discord than anything else in 1-2 weeks.

...all keet is good for is to get a quick impression of whats possible with the dat stack.

here are screenshots from the students project.

@pospi @strypey

dat ecosystem is volunteers only.
unlike for profit often VC funded companies, dat is 501c3 non-profit public good company with code of conduct and all the proper commons community based infrastructure.

the dat stack started in 2013 is the most advanced most battletested and mature stack in the history of humanity - sounds dramatic but its actually the facts.

ppl could support and build amazing apps,

...but if nobody supports dat with funding or volunteer work - then...

@pospi @strypey
.. of course we will be stuck with scam crypto and VC funded enshittification ware, who always brush apps to look capable and amazing so everyone adopts them (e.g. see bluesky) ...or even mastodon (where funding comes from big government and in practice things centralized around mastodon.social) ...by design basically.

ppl you need to make up your mind and at least not actively fight the few that try to push things forward against all odds.

@strypey

dat is a p2p stack that is open source, not a ready made app.

maybe try mapeo then ... but mapeo is an offline first local geomapping app.

you cpuld probably also try cabal chat.

i sent you a series of urls to dat based apps, but still to get a feeling for better or worse, keet is the most polished app when it comes to ui/ux.

i'm working on a dat web based rss reader + blogging app, but its not officially released yet.

here is a preview:
https://youtu.be/NszqD5ZipY4

0 installation

Worklog 14 - UI Demo of subscriptions sync event_drive

YouTube

@serapath
> [HoloChain] have VCs

Yes, taking VC was a mistake, as it was for GL, Matrix, etc, etc.

> tokensale to fund their token on ethereum

Links please? This is not my understanding of what they've been up to.

> imho firmly part of the web3 scam

Maybe. But I met a number of core HoloChain devs at the Open conference in 2018, some of them sat in on my open session;

https://inv.nadeko.net/watch?v=WJR7IHeLADo

They seemed pretty genuine about trying to decentralise hosting with Free Code and open protocols

@strypey

yes, that is true.
i also know nice holochain devs like victor saunders.. i dont think that many of them know or really are in touch with the VCs, CEOs, etc... of thr folks who run the show, bit they just build and use the tech - which is exactly what their marketing intends.

just search for ICO, HOT token launch and their HoloHost ...imho a spooky name, but yeah.

@serapath
> i dont think that many of them know or really are in touch with the VCs, CEOs, etc... of thr folks who run the show, bit they just build and use the tech

I'm not talking about people building apps on HoloChain. I'm talking about ...

> core HoloChain devs

The people who were developing HoloChain from scratch at the time.

> search for ICO, HOT token launch and their HoloHost

I'm familiar with Holo.host. I don't think it works the way you're describing. Link, or retract the claim.

@strypey

I dont understand what you are asking.
Imho in 2018, they launched the HOT token on ethereum to raise ~20 million in an ICO and thats how they fund stuff.

I think thats public information.
I just read they integrated iroh tech, which is a new york based for profit company that where key devs came feom protocol labs building IPFS, another venture capital funded company.
Seems holochain integrated iroh now...

They tag themselves with #web3 and #dapps ...
https://coinmarketcap.com/community/profile/Holo/

Follow The Crypto Message Board On Community

Join in the conversation on CoinMarketCapโ€™s crypto message boards today. See comments and alerts posted by leading projects and people in cryptocurrency.

@serapath
> you could probably also try cabal chat

This one?

https://cabal.chat/download.html

https://github.com/cabal-club

Last commit in 2024. They started developing their own protocol;

https://github.com/cabal-club/cable

... then dev seems to have stalled. This is what I'm talking about. The pure P2P world is full of idealistic goals, bold claims and flashy websites, but scratch the surface and it never. Works. Out.

This is why I can't be bothered trawling through lists of links and asked for 1 recommendation.

cabal - downloads

p2p chat

@strypey

cabal started building their own protocol to replicate data, but they continue to use dat for peer discovery to sync.

maybe focus on mapeo then, but i dont think you are interested in actually trying. it costs nothing to install keet, check it out and then delete it. ๐Ÿคทโ€โ™€๏ธ

anyway - unlike other VC funded proejcts, dat is volunteers and 501c3 non profit public good company with a manifesto and transparent governance.

...yes. for profits have an easier time to make nice things. ๐Ÿคทโ€โ™€๏ธ