Hacker News Reader

Comment by kodablah | Hacker News Reader

I like the idea of Ethereum but it does not solve distributed application needs IMO. I fear it is not able to solve large-scale app problems considering it only has the computation power of a "1999 phone", requires every to store (or at least know about) all state, and requires everyone up to date to run EVERY computation. For instance, take a distributed Reddit where posts can be mappings, even by deferring the costs to the viewers in terms of small amounts of wei per view, it surely wouldn't be enough to cover the expense of centralized permanent store (I understand Whisper is coming but it isn't permanent enough).

I see a couple of Ehtereum frameworks out there using things IPFS for storage. Are there any plans for Ethereum to solve concerns about forcing everyone to run all calculations or should it always be seen as a small DHT that you only want to use when you have real auditability concerns?

View on HN

This is probably the most cogent criticism that can be leveled against ethereum right now- The performance constraints for ethereum dapps are pretty extreme. There are indeed plans to solve the problem you describe (read posts on blog.ethereum.org mentioning scalability) and these problems are probably solvable, but a working system remains far off.

That said, I think ethereum dapps are going to be substantively different from traditional apps people are used to- I think there's a lot of opportunity in this type of system for people who don't take the mindset of "This system is bad at running existing app X therefore it is bad for everything."

I think criticism is valid in regards to TPS. But for storage it's misguided. The blockchain is meant for consensus, storing reddit comments or cat pictures is an expensive and inefficient misuse of the blockchain.

Hashes in the blockchain, when necessary, and then markets which will sell you a file when presented with the hash. Storage technology can be whatever.

That's the smart plan, I think. I really want to take a crack at those storage markets in-browser, use a distributed hash table and WebRTC to pay people for leaving a browser open and allowing us to store stuff on their hard drives!

36 cents will get you a gigabyte stored for a year with 99.999999999% durability and 99.99% availability at S3, accessible at multiple gigabits per second.

When gigabit class connections are pervasive, the idea of paying people to store data in an open web browser is maybe feasible, but certainly not profitable for any involved party.

Before then? I dont even see the point.

Ethereum doesn't aim to store massive data within the blockchain itself. Instead, we're building an additional component into the network, Swarm, which essentially acts as a DHT/DFS. There are already experimental branches within the main repos, they function, but probably at a POC state.

Furthermore, a 1Gbit DL/200Mbit UL link costs $13.70/month where I live, so I think it it is already very much viable, but maybe the distribution will be skewed towards more internet friendly countries :)

True it might be misguided for Ethereum, but I meant my comment to be taken in the context of other distributed crypto tech being worked out. It was more a criticism of the blockchain. Surely we can find some middle ground for a "web-of-trust" between a few nodes and everyone. I think maidsafe settled on 32...I forget, and I don't know enough to know the dangers there.

But yes, I realize that for storage it is ideal to use another service, I just fear normal developers inability to tie the two safely...frameworks like Embark do seem help there (though I find IPFS too slow for my stateful needs).

Ethereum is indeed not meant for distributed computing, You should use Ethereum only for consensus and avoid any unneeded processing, the blockchain is there to avoid 'cheating' it's not meant to be a backend.

As for frameworks I guess you are referring to Embark: https://github.com/iurimatias/embark-framework

Agreed. Right now it's a single shared processor emulated with cryptography and massive replication - but that's not an architecture that will get you supercomputer performance.

The work on that thread - getting Ethereum to supercomputer status - is called CASPAR and you can read more about the efforts to get a formal proof for it here:

http://blog.synereo.com/2015/07/23/Ethereum-+-Synereo/

That's because you're looking at the platform with the wrong perspective...

When the Uber app needs consensus with the banks in order to mark a payment as "legit", it uses just few kb's of bandwidth and a couple of requests to communicate with Braintree. You don't run all your apps on Braintree's shoulders.

The same way, I can build a normal app, or NW.js wrapper for my desktop app (HTML/JS/CSS) and use localstorage as db, or remote servers as backends to process the heavy tasks. Then I'd connect to the Ethereum network just when is required consensus. I don't have to run all the app on Ethereum.

That's how I see it. You don't use a blockchain as a DB to store Reddit posts. It's a poor choice. Ethereum is very good for other use cases.

Reply on news.ycombinator.com