Ethereum: 2nd gen cryptocurrency with contract programming, "dagger" hashing

[tacotime]

Interesting project with a novel algorithm. From what I understand, the algorithm (Dagger) is memory-hard not necessarily in bandwidth, but in space; using something like a GPU (with many threads) would benefit most from having lots of system memory:

This algorithm provides a proof of work mining function with memory hardness properties that are not ideal, but that are nevertheless a massive improvement over anything available previously. It takes 512 MB to evaluate, 112 KB memory and 4078 hashes to verify, and even the tinest time-memory tradeoff is not worthwhile to implement because of the bottom-level branching adjustment. These parameters allow Dagger to be much more daring in its memory requirements than Primecoin or scrypt, asking for 512 MB of RAM for a single thread. Because the primary determinant of hardness is memory, and not computation, specialized hardware has only a tiny advantage; even an optimal Dagger mining ASIC would have little to offer over a hobbyist purchasing hundreds of gigabytes of memory cards off the shelf and plugging them into a medium-power GPU. And even in such an equilibrium, mining with ordinary CPUs will likely continue to be practical.

512MB per thread .. wow.

I'll remain skeptical of possible speculation about the absence of TMTO solutions, as I've certainly heard this one before.

[tacotime]

I'm not knowledgeable about how fast GDDR5 is vs. DDR3, but I don't think it's an order of magnitude right?

About half an order of magnitude in terms of BW

The main problem with vRAM tends to be latency, but it's less of an issue if you plan your implementation well

[romerun]

i'll mine this coin

[tacotime]

What I don't understand is who will do the computing for the Turing-complete "contracts"?

So a "contract" is supposed to be some script that is embedded on the blockchain. People can send messages to the contract that it will take as input to some computation. Maybe a simple example would be a vending machine. I send some bitcoins with a message "I want a Coke" to the vending machine contract. The vending machine processes my input and directs the physical vending machine in front of me to vend a Coke.

But that code has to run on a computer somewhere. Does it run on a miner's computer? If that's true then what prevents me from injecting malicious code? For example code that never halts and ties up resources?

Even if you don't have malicious code I could imagine people building complicated scripts that might tie up computation resources. How do you create an effective incentive scheme for miners then? You can't just use the code size because there is no relation between code size and resources required to run the code.

It looks like they try to mitigate this by taxing operations with fees:

Code:

TXFEE (100x) - fee for sending a transaction
NEWCONTRACTFEE (100x) - fee for creating a new contract, not including the memory fee for each item in script code
STEPFEE (x) - fee for every computational step after than first sixteen in contract execution
MEMORYFEE (100x) - fee for adding a new item to a contract's memory, including when first creating a contract. The memory fee is the only fee that is not paid to a miner, and is refunded when memory from a contract is removed.
DATAFEE (20x) - fee for accessing or setting a contract's memory from inside that contract
EXTROFEE (40x) - fee for accessing memory from another contract inside a contract
CRYPTOFEE (20x) - fee for using any of the cryptographic operations

[GigaCoin]

I was aware of this since about a week from the i3 forums. Apparently creator of ethereum is an ex founder of invictus.

I also read the white paper seems promising, however I see no mention of an IPO like OP says or am I missing something?

[simbo]

Very interesting!!! 

[jimhsu]

You could even put sha256 mining code in your contract. So anyone can come along and contract an Ethereum miner to mine bitcoin for them.

That is ... genius. The thought of an idea like that actually being possible.

[Ghepetto]

"a superior foundational layer offering a uniquely powerful scripting system on top of which arbitrarily advanced contracts, currencies and other decentralized applications can be built"



Is this similar to the madesparq project? 

[tacotime]

I was aware of this since about a week from the i3 forums. Apparently creator of ethereum is an ex founder of invictus.

I also read the white paper seems promising, however I see no mention of an IPO like OP says or am I missing something?

There's supposed to be a mastercoin style IPO where a certain number of USD/BTC yields some ETH

[canth]

Can it be mined with GPU?

Yes and no. Dagger has a minimum of 512MB memory per thread, so while it would be GPU mineable, it doesn't appear that this would be superior to CPU mining. It looks like blade servers with a 512MB/core ratio would be the most effective miners, at least off the bat.

[Chang Hum]

one to watch

[SyRenity]

I would speculate from the at least 3-5 people currently hacking on GitHub (and the rumoured current staffing of 5-20 people), that they already have a massive amount of funding running into the millions of dollars.

This gives a Ripple de-javu. These millions have to be paid off some way, any chance it will be 90% for Eth. and 10% for the community?

[miramare]

Really a very interesting project.

I'm looking forward to more info.

[charleshoskinson]

Alright guys since you're speculating about Ethereum, I figured I'd drop by and say hi. I'm one of the core developers. Go ahead and post your questions here

[atleticofa]

It looks a great project!

[td services]

Will the forum be back up or is there a new link to it?

[tokyoghetto]

Alright guys since you're speculating about Ethereum, I figured I'd drop by and say hi. I'm one of the core developers. Go ahead and post your questions here

[charleshoskinson]

Will the forum be back up or is there a new link to it?

We've been mostly trying to run in silent mode for a bit, but the project got too big too quickly and thus the flood of traffic crashed our website. We should have a better foundation in play and the blog, wiki, forums as well as website fully operational by the end of the week. We have a hard limit for the Miami conference.

We are also getting a newsletter setup alongside some partner integrations to help us both with the fundraiser and keeping the community informed about what's going on. Believe me when I say that what you have seen is the very tip of a very large iceberg.

[SyRenity]

Alright guys since you're speculating about Ethereum, I figured I'd drop by and say hi. I'm one of the core developers. Go ahead and post your questions here

I would speculate from the at least 3-5 people currently hacking on GitHub (and the rumoured current staffing of 5-20 people), that they already have a massive amount of funding running into the millions of dollars.

This gives a Ripple de-javu. These millions have to be paid off some way, any chance it will be 90% for Eth. and 10% for the community?

[charleshoskinson]

This gives a Ripple de-javu. These millions have to be paid off some way, any chance it will be 90% for Eth. and 10% for the community?

While we as a project have not accepted a single bitcoin in funding yet, there are some significant commitments. All fundraising will be done in an open and transparent way with equal treatment for everyone.

We are currently exploring the notion of a BTC to Ether swap and thus this would not infer control or voting rights to the investors in any way. The only way anyone (including the founders) make money with Ethereum is if it realises the utility the whitepaper has outlined. I believe with the team we have and the needs of the market, this will be the case on a grand scale.

I would like to add, once the website is up, I think everyonemost people(troll insurance) will be very happy with the distribution model. We spent a lot of time trying to model it after a startup as it grows to a full IPOed company. We are not like Ripple nor Bitcoin nor protoshares. Ethereum is its own model.