Re: Tau-Chain and Agoras Official Thread: Generalized P2P Network

Any news on the project?

I'm still wrapping my head around converging to the best compilation architecture for tau. It is a delicate and hard task.
BTW this is from today: http://bitcoinist.net/six-ethereum-projects-and-its-five-competitors/

You've been working on that for a long time. Have you considered working on the actual code to bootstrap Tau first, and come back to this later?

I know that when building something as complex as a new language, and pieces have to work together, but once you have the code that the Tau engine will compile, it will be a lot easier to create an efficient compiler, "A plan never survives first contact with the enemy", "Make it work, make it right, make it fast" and all that...

Also, having something to show, as slow and imperfect as it may be, would do well for getting the message out and help you raise funds. This as well as the effect on the community and helping draw smart people to be involved in the product, which I would expect to be invaluable...

I agree...

I really wish it was possible (beginning tau network without the 'best' compiler and finishing it later), and since the beginning we gave a decent thought to 'what is the bare minimum'. Obviously that'd make 'life much easier' for me.
But the execution path must be predetermined at genesis and beyond, so given a proof by one client, it can be verified by another client - at the very same proof flow. If we would give up rigid and predetermined proof flow, we won't be able to implement lambda-auth (hashing the proof tree in a way that leaves a short 'proof of correct proof' or 'proof of execution', relying on the fact that all clients have exact same flow).
So this underlying flow cannot be changed and must be finalized at the best possible performance. Which is nothing but the compilation architecture.

But wouldn't that leave room for a test network? A test network wouldn't require the best possible performance, and the code (logic? I'm not sure what to call the stuff that makes up the Tau network) should run as is on production once you get there, right? And that way, you could assign the creation of the code running on Tau off to some other people, because it would be mostly orthogonal, isn't it? And also, people could independently try and improve your engine/create their own, since they'd have a baseline to run theirs against.