[gonehome]

For those that haven’t checked it out the ideas behind urbit are really interesting, it’s the first attempt I’ve seen to get us out of the local maximum of centralized web services we’re trapped in where success is among the potential outcomes (most attempts are destined to fail because they don’t tackle the underlying issues from first principles). I think it could really work and I’m convinced urbit or something like it eventually will - the ideas are correct.

Jack Dorsey’s “web5”, Dan Romero’s Farcaster all touch on a lot of these concepts - people are starting to come around to the same conclusions (which is good!) - urbit had just taken a boil the ocean approach on a bet that we need to rethink the stack from first principles so we can control certain assumptions to reduce complexity and make some of the problems in modern computing easier to handle.

Finding it really felt like the first time I started playing with Linux for me. It revived a lot of the dreams of the 90s web and what it could be, people being able to run their own nodes in a way that could actually work.

To do that you have to solve the identity problem, you have to reduce the complexity of running the system, and you have to have the internet front and center when building the OS.

The dream of a truly decentralized web from the 90s failed and it failed for reasons that make sense given the constraints of the technology we use. I think urbit is a way out.

An old blog post (from 2010!) that really digs into the details: https://moronlab.blogspot.com/2010/01/urbit-functional-progr...

[cowtools]

I've looked into it and it sounds like a ton of pretentious crap. How is this any better than IPFS, I2P, tor onion services, or Gnunet?

It's nice that it helped introduce you to linux though.

You make a few good points about usability, but I don't think the internet was designed to be a distributed network from the get-go. https://secushare.org/broken-internet

[gonehome]

It didn’t help introduce me to Linux - I used Linux for 15yrs before I heard about urbit.

I was disappointed that Linux would never be able to solve the incentive problems that lead to centralized companies handling that complexity.

Urbit’s approach actually could.

The reasons those other attempts are basically DOA outside of a tiny technical niche is they don’t go deep enough into the stack, they don’t tackle the issue with identity which is a core problem when it comes to handling spam and moderation. They don’t tackle administrative complexity and dependencies which is a core issue in making the system simple to run.

To really fix these things you have to think about why they always fail and design something to fix those incentive issues. Those problems start earlier up the stack - Linux will never be able to do it (imo) even if you tack on the identity layer.

[cowtools]

GNUnet is a pretty comprehensive solution that does address these things, albeit it is a work in progress. Their solution to the identity problem is a pretty ingenious petname system called GNS. Their solution to incentive issues is a cryptographic micropayment system called GNU Taler. I think you can pretty easily solve the problems "up the stack" if you make the systems transport-layer agnostic like GNUnet/IPFS has done, and like Tor/I2P are currently doing.

I think Linux's identity solution (if you consider it a solution) is GPG and PKI.

[gonehome]

I’ll look at GNUnet - I don’t know enough about it to comment.

GPG is largely a failure imo, I think Moxie is right about this - complexity is high (even when people’s life depends on it they won’t use it) and the core issue is that it’s distinct from the system itself.

Urbit nodes cannot be separated from the ID - you can’t run it without an ID which is a pretty clean way to solve a lot of this stuff out of the box (also adds a lot of cool features/guarantees you can make about collaborating with users over the network, ability for pseudonyms to accrue reputation, easy ability to moderate, and make spam not economically possible, etc.

I’ve become convinced you can’t solve these things separately in parts on the Linux stack, maybe someone will - but there’s been 30yrs of failure so I’m not optimistic.

[arthyn]

https://news.ycombinator.com/item?id=9224

[gaze]

It's in no way interesting. It's a bunch of old, bad ideas duct taped together and renamed with nonsense words to both obfuscate what they're doing and establish an in-group language just as cults do.

[0x732202]

Binary trees as a fundamental data structure is not a bad idea, and it's also entirely clarifying for programming purposes. Don't you agree OOP has huge drawbacks?

[vorpalhex]

I would like to call out that parent here is not a neutral or random party and works on Urbit, and so may have a conflict of interest.

[gonehome]

Yeah it’s not a secret (it’s in my bio), I left to join Tlon at the start of this year but found urbit a couple of years ago. The comments here are similar to ones I’ve made on HN before (prior to working on it full time).