17 comments

[ 2.5 ms ] story [ 46.8 ms ] thread
I came across this project years ago and just rediscovered it today. I think it’s interesting how they focus on networking from the perspective of IPC and build the system by layers different implementations of the same interface
This looks like a student assignment project from some CS class. A lot of promises, but looks like some TCP reinvented over UDP using some C API with some decentralization marketing.

> TCP/IP design is fundamentally broken, and a different approach is needed to build efficient packet networks.

I wonder how it compares to QUIC
They support doing this over ethernet packets as well; they don't require IP networking. Supporting IP/UDP is a sane thing to do for supporting legacy networks.
So what you're saying is better than most startups? At least this one has a naive technical premise rather than "disrupting [non-existent problem]"
FYI: CTA buttons can't be clicked when they're behind the transparent header, i.e. when you've scrolled down enough that the buttons go where the header is.
> Ouroboros congestion control operates at the network level. It does not (can not!) rely on acknowledgements. This means all network flows are automatically congestion controlled.

How exactly does this work? Once you send a packet into the network, you have no way of knowing whether or not it is received by the recipient, unless the receiver sends an acknowledgement - and the only symptom of congestion is dropped packets, i.e. recipient not receiving a packet..

And a follow-up, how is that different than what UDP solves?
Interesting concept, but a more modern, more secure set of tools at this level these days should be done in a language that makes secure coding easier than C. I get they want to easily offer a C-level API, but at least use something that machine-generates correct C.
> This immutability can be enforced through authentication (not yet implemented).

So, that's one of the most interesting parts...