nat traversal more or less solved:

https://docs.iroh.computer/concepts/nat-traversal

https://www.iroh.computer/blog/what-is-derp

https://ark-builders.medium.com/the-deceptive-complexity-of-...

projects using iroh:

https://github.com/muhamadjawdatsalemalakoum/kith

https://delta.chat/

https://github.com/fishfolk/bones/

I'm pleased to see DAT survived in some way the niche rivalry with IPFS, but let the people at Iroh their website LLM endpoint is down, https://www.iroh.computer/blog/what-is-derp

> "The DERP Protocol in Iroh"

> "Explain the original DERP protocol by Tailscale, focusing on its functionality and limitations. Discuss how Iroh has adapted and enhanced the DERP protocol, emphasizing the use of DERP URLs instead of regions."

and so on for the rest of their FAQ page

PS: Is keet.io a reincarnation of Cabal chat (https://cabal.chat/)?

There's no LLM endpoint :) These prompts were meant for humans to fill in :P (this post was from 2024)

That said, this was an old blog post that never actually got finished, so we've taken it offline and replaced it with a redirect to our relay docs page.

no. cabal chat is different.

...but you can also checkout quibble as an alternative to keet that is open source.

https://codeberg.org/LTC-GT/Quibble

i think the most awaited thing that has been in the making for hlf a decade now is "datweb" so all the dat ecoaystem projects might potentially work in browsers as well and not just as standalone apps

nat traversal decentralized/distributed holepunching is absolutely what dat has pioneered and is also industry leading and battle tested because it has plenty of apps with lots of users.

dat exists in rust and the more feature rich mature version is C & JS so you cqn build cross platform p2p apps using the web stack to make it as easy as possible for anyone to adopt

Falling back to DERP relays means trusting and funding centralized relay boxes. That solves connectivity by conceding peer-to-peer. You end up with client-server architecture plus extra cryptographic tax.

There's a distinction to be made between p2p as a network topology and p2p as a redistribution of power and agency. It's true that relays are an unsatisfying fallback to deal with the networks we have, but they still allow for meaningful p2p in sociopolitical terms. Relays that support the spirit of p2p need to be federated, stateless, and end-to-end encrypted.

you dont need relays if you use dat.

blind relays are optional and literally only relevant for specific really difficult carrier grade network conditions where direct connections might not always work.

...but even then, relay is really dynamic and any dat peer is automatically and dynamically participating im the process ..p2p, not federated

Are you saying that, aside from carrier grade NATs, dat otherwise achieves 100% direct connections?

i mean - network situations can be complicated and problematic but i would say pretty much yes in the sense of in prwctice it almost always works - 100% is too strongly framed, but it is really reliable, feel free to try.

if you dont want to use the stack and try programmatically, try keet messenger or holesail.io app to experience the reliability

Paul Frazee, maker of the late Beaker Browser, a browser with p2p via Dat, started and is CTO at Bluesky, doing atproto. https://en.wikipedia.org/wiki/Beaker_(web_browser)

They have a great talk about solving for scale. I feel like atproto has really tackled so many of the incredible challenges of getting p2p to work, by honing in on exactly these sorts of questions. https://www.youtube.com/watch?v=BoJnj2yPf14

atproto is not p2p and will never be. ...if thats why ppl adopt it, they essentially got scammed.

p2p has lots of challenges but many are cultural and social and related to funding and changes in UX, but the technical side is pretty much battle tested and production ready and working.

again, paul pfrazee is the creator of beaker browser - one dat ecosystem project among many, so he is of course entitled to his opinion

Just pick the right technologies for the right problems. If your scale is entirely interpersonal - small orgs, friends, family devices - then p2p is much more likely to be a good choice.

at any point in time, a person can only read so much scrolling through their feed and follow thousands.

try keet.io and join rooms woth 20k+ members. works without problems.

even having millions of followers isnt a different problem to torrenting a popular archive.

p2p absolutely scales and excels when the amount of users is massive.

"... never escape the hobbyist tier."

https://en.wikipedia.org/wiki/Homebrew_Computer_Club