5 ms·

Scalability, Part 2: Hypercubes

55 points by jrbedard 12y ago

throwawayether 12y ago
Love to see a poll: will Ethereum ever launch? I can't tell if this is all horseshit, or too deep for me to understand.
- wyager 12y ago
  >I can't tell if this is all horseshit, or too deep for me to understand. A mixture of horseshit and mildly interesting stuff. Even though ethereum is technically feasible, it's plagued by developmental issues (including, IIRC, failed attempts at roll-your-own crypto), and even if it does ever get launched, the incentive structure doesn't make sense and the whole platform is essentially useless. It's absolutely no good for distributed apps; you still need oracles for those to do anything useful, and if you have oracles, you might as well put the apps on the oracles in the first place. It might be useful for digital contracts, but Bitcoin already handles those very well in the vast majority of cases.
  - DennisP 12y ago
    It's not like they're coming up with their own encryption algorithms. The original parts are things that haven't been done before. Roll-your-own is the only option. It's worth noting maybe that Ralph Merkle is on the team. http://www.reddit.com/r/Bitcoin/comments/25269r/dr_ralph_merkle_of_merkle_tree_fame_just_joined/ http://www.reddit.com/r/Bitcoin/comments/25269r/dr_ralph_mer... The only thing you need oracles for is getting information from the outside world. I can think of plenty of applications that don't require that. But even then you don't really need oracles if you use the Schelling mechanism they've been talking about.
    - wyager 12y ago
      >It's not like they're coming up with their own encryption algorithms. I seem to remember something about a tree-based hash algorithm for mining purposes that turned out to be cryptographically insecure. >The only thing you need oracles for is getting information from the outside world. Or for having any effect on the outside world (i.e. the things that people usually want programs to do). > I can think of plenty of applications that don't require that. Besides financials/contract processing, what? >Schelling mechanism Do you have a link about this?
      - DennisP 12y ago
        Schelling: https://blog.ethereum.org/2014/03/28/schellingcoin-a-minimal-trust-universal-data-feed/ https://blog.ethereum.org/2014/03/28/schellingcoin-a-minimal... https://blog.ethereum.org/2014/06/30/advanced-contract-programming-example-schellingcoin/ https://blog.ethereum.org/2014/06/30/advanced-contract-progr...
- semiel 12y ago
  I've been following development pretty closely, and I'm reasonably convinced that they will launch in the first half of 2015. I wouldn't be surprised if they missed their target ("before March 21st") by a month or two, but they seem committed to getting a reasonable 1.0 out the door soon. This was much less clear a few months ago, since they kept coming up with yet another company-sized idea to implement every week or so. But apparently Nick Szabo put the fear of god into them and reminded them that there's no point in designing a perfect system that never ships. It's worth noting that this blog post is explicitly about ideas for the as-of-yet hypothetical version 2.0. They don't need to figure this out in order to launch.
bkirwi 12y ago
As a side project, I've been reimplementing a large chunk of Ethereum PoC-5 in Haskell. Given all the stuff that's going on in that ecosystem, though, I've basically stopped development. Bitcoin is held together by a delicate balance of cryptography, software design, and community involvement. Most altcoins are very conservative about the changes they introduce, keeping them as close as possible to successful designs and carefully monitoring the blockchain for stability. Ethereum, meanwhile, no longer bears much resemblance to the projects that inspired it. Even if Ethereum's core algorithms are sound -- and it's becoming difficult for anyone outside the project to evaluate that at this point -- there's no concrete evidence that it will be stable in practice, and they're already building many layers of code on an uncertain foundation. I'd love for Ethereum to succeed; I've always been more interested in Bitcoin as a distributed system than as a currency, so it's great to see people pushing it in that direction. But it's gone past the point where I can evaluate all the new stuff that's happening, and I feel weird devoting time to a project when I can't see where it's headed. We'll see what happens once the dust settles...
- semiel 12y ago
  Personally, this is why I find Ethereum so fun. I don't think anyone expects Ethereum to suddenly take over the world, but it's a really interesting testing ground for a lot of fascinating ideas. If we were trying to identify a spiritual predecessor, it would definitely not be something like the Apple II. We're much more in "Mother of All Demos" territory. I would honestly be mildly surprised if I was actually making money programming Ethereum contracts in 10 years, but I would not be remotely surprised if I was doing something that owed its existence to the experiments the Ethereum team is doing.
- DennisP 12y ago
  The article here is a version 2 feature. They're planning to stay a little more down to earth with the first release. Hopefully that will keep them from screwing up too badly.
Animats 12y ago
Determining if that algorithm is secure against attacks which steal value is going to be tough. Even if the algorithm is secure, the implementation may not be. Building a general-purpose interpreter into a money system may not be a great idea from a security perspective. Could someone put hostile code in the block chain? Cause overflows? I don't know. This is a really complicated consistent-eventually system. It's possible for parts of the network to be out of sync, possibly due to attacks. Eventually, the inconsistency is discovered when two out-of-sync subnets communicate. Then there's a complex dispute-resolution strategy which ultimately results in some blocks being invalidated. It's not clear how long this takes. Since the application is money transfer, a substantial delay may offer the opportunity to take the money and run, which is almost standard practice in the Bitcoin world. (A long, long time ago I looked at a similar problem - keeping routing tables in sync in the presence of bogus data insertion. I never had a really airtight solution; the best I was able to do was to come up with a way to detect attacks. If there's now a good solution to that problem, perhaps it should go into a successor to Border Gateway Protocol, which has route forgery problems.) Checking out the security of this thing looks like a PhD thesis sized project, with a really good crypto advisor.
- bhouston 12y ago
  > This is a really complicated consistent-eventually system. So it is very complex and it is supposed to be trusted for monetary transactions, and failure means significant financial loss? This isn't a good combination to have. It is great to have these types of really complex ideas, but something more elegant is likely to beat out this idea in the long run.
- DennisP 12y ago
  Well then it's good that they have a team of people working full-time, and Ralph Merkle.

55 points by jrbedard 12y ago

throwawayether 12y ago

Love to see a poll: will Ethereum ever launch? I can't tell if this is all horseshit, or too deep for me to understand.

wyager 12y ago
>I can't tell if this is all horseshit, or too deep for me to understand. A mixture of horseshit and mildly interesting stuff. Even though ethereum is technically feasible, it's plagued by developmental issues (including, IIRC, failed attempts at roll-your-own crypto), and even if it does ever get launched, the incentive structure doesn't make sense and the whole platform is essentially useless. It's absolutely no good for distributed apps; you still need oracles for those to do anything useful, and if you have oracles, you might as well put the apps on the oracles in the first place. It might be useful for digital contracts, but Bitcoin already handles those very well in the vast majority of cases.
- DennisP 12y ago
  It's not like they're coming up with their own encryption algorithms. The original parts are things that haven't been done before. Roll-your-own is the only option. It's worth noting maybe that Ralph Merkle is on the team. http://www.reddit.com/r/Bitcoin/comments/25269r/dr_ralph_merkle_of_merkle_tree_fame_just_joined/ http://www.reddit.com/r/Bitcoin/comments/25269r/dr_ralph_mer... The only thing you need oracles for is getting information from the outside world. I can think of plenty of applications that don't require that. But even then you don't really need oracles if you use the Schelling mechanism they've been talking about.
  - wyager 12y ago
    >It's not like they're coming up with their own encryption algorithms. I seem to remember something about a tree-based hash algorithm for mining purposes that turned out to be cryptographically insecure. >The only thing you need oracles for is getting information from the outside world. Or for having any effect on the outside world (i.e. the things that people usually want programs to do). > I can think of plenty of applications that don't require that. Besides financials/contract processing, what? >Schelling mechanism Do you have a link about this?
    - DennisP 12y ago
      Schelling: https://blog.ethereum.org/2014/03/28/schellingcoin-a-minimal-trust-universal-data-feed/ https://blog.ethereum.org/2014/03/28/schellingcoin-a-minimal... https://blog.ethereum.org/2014/06/30/advanced-contract-programming-example-schellingcoin/ https://blog.ethereum.org/2014/06/30/advanced-contract-progr...
semiel 12y ago
I've been following development pretty closely, and I'm reasonably convinced that they will launch in the first half of 2015. I wouldn't be surprised if they missed their target ("before March 21st") by a month or two, but they seem committed to getting a reasonable 1.0 out the door soon. This was much less clear a few months ago, since they kept coming up with yet another company-sized idea to implement every week or so. But apparently Nick Szabo put the fear of god into them and reminded them that there's no point in designing a perfect system that never ships. It's worth noting that this blog post is explicitly about ideas for the as-of-yet hypothetical version 2.0. They don't need to figure this out in order to launch.

bkirwi 12y ago

As a side project, I've been reimplementing a large chunk of Ethereum PoC-5 in Haskell. Given all the stuff that's going on in that ecosystem, though, I've basically stopped development. Bitcoin is held together by a delicate balance of cryptography, software design, and community involvement. Most altcoins are very conservative about the changes they introduce, keeping them as close as possible to successful designs and carefully monitoring the blockchain for stability. Ethereum, meanwhile, no longer bears much resemblance to the projects that inspired it. Even if Ethereum's core algorithms are sound -- and it's becoming difficult for anyone outside the project to evaluate that at this point -- there's no concrete evidence that it will be stable in practice, and they're already building many layers of code on an uncertain foundation. I'd love for Ethereum to succeed; I've always been more interested in Bitcoin as a distributed system than as a currency, so it's great to see people pushing it in that direction. But it's gone past the point where I can evaluate all the new stuff that's happening, and I feel weird devoting time to a project when I can't see where it's headed. We'll see what happens once the dust settles...

semiel 12y ago
Personally, this is why I find Ethereum so fun. I don't think anyone expects Ethereum to suddenly take over the world, but it's a really interesting testing ground for a lot of fascinating ideas. If we were trying to identify a spiritual predecessor, it would definitely not be something like the Apple II. We're much more in "Mother of All Demos" territory. I would honestly be mildly surprised if I was actually making money programming Ethereum contracts in 10 years, but I would not be remotely surprised if I was doing something that owed its existence to the experiments the Ethereum team is doing.
DennisP 12y ago
The article here is a version 2 feature. They're planning to stay a little more down to earth with the first release. Hopefully that will keep them from screwing up too badly.

Animats 12y ago

Determining if that algorithm is secure against attacks which steal value is going to be tough. Even if the algorithm is secure, the implementation may not be. Building a general-purpose interpreter into a money system may not be a great idea from a security perspective. Could someone put hostile code in the block chain? Cause overflows? I don't know. This is a really complicated consistent-eventually system. It's possible for parts of the network to be out of sync, possibly due to attacks. Eventually, the inconsistency is discovered when two out-of-sync subnets communicate. Then there's a complex dispute-resolution strategy which ultimately results in some blocks being invalidated. It's not clear how long this takes. Since the application is money transfer, a substantial delay may offer the opportunity to take the money and run, which is almost standard practice in the Bitcoin world. (A long, long time ago I looked at a similar problem - keeping routing tables in sync in the presence of bogus data insertion. I never had a really airtight solution; the best I was able to do was to come up with a way to detect attacks. If there's now a good solution to that problem, perhaps it should go into a successor to Border Gateway Protocol, which has route forgery problems.) Checking out the security of this thing looks like a PhD thesis sized project, with a really good crypto advisor.

bhouston 12y ago
> This is a really complicated consistent-eventually system. So it is very complex and it is supposed to be trusted for monetary transactions, and failure means significant financial loss? This isn't a good combination to have. It is great to have these types of really complex ideas, but something more elegant is likely to beat out this idea in the long run.
DennisP 12y ago
Well then it's good that they have a team of people working full-time, and Ralph Merkle.