I think they are mostly waiting for MaidSAFE to solve that problem, which is a group that's actually serious about the intricate technology that needs to go into this, and all the special cases (e.g. how many times do you need to replicate a chunk of data to make sure it isn't lost when people turn off their computers). See https://github.com/maidsafe
I've been mostly unimpressed with MaidSAFE approach to the storage problem.
They (last I checked) are approaching the problem by using quadruple redundancy, no erasure coding. This opens some attack vectors, because an attacker only needs to control 4/4 pieces, something that takes a probability of (attacker control)^4 . Their proof of storage also seems to use nonce's, which require precomputation before uploading the file, and means a limited amount of proofs before you have to fetch the file and compute a bunch of other nonce's.
The biggest problem with MaidSAFE's stuff is that it's all over the place. There's no concise document that explains proof-of-storage. Their approach seems enormously complex, which to me indicates a huge likelihood of undetected vulnerabilities (complexity is bad when you want security).
That paper was written before decentralized storage was really just a new idea. I've since implemented a baseline core algorithm, and we have one application running.
I'm in the process of rewriting the whitepaper. Hopefully will put out a draft in two weeks(I'll be at conferences all this week). From there I want to actually collect network data(after all I'm trained as a researcher and what is good paper without any data?) before putting out a solid draft to be peer reviewed.
I really couldn't find any precise information about storj but the dummy slides are showing some really different scheme for POR.
Also I think filecoin has a mistake. If it ever gets popular then it is possible that the difficulty would be quite high (in the scale of GBs). In that case you can't effectively mine only if you share at least 1 TB of your hard drive below that the expected reward for a block would be near 0. It would drive the network to centralization. The expectation reward for a block isn't linearly proportional to your shared disk space but a convex function of it.
On the other hand it would be possible to implement mining pools which could somewhat solve this problem.
I'm working on a highly distributed coop cloud built with OpenStack and Bitcoin: https://www.stackmonkey.com/. The appliance is in beta as of yesterday, but the site is still in progress.
I'm at a slightly different layer of the stack - one that might be advantageous for users wanting to run decentralized offerings like Storj and Maidsafe without having to do the deployments themselves. If any of the Storj guys are interested, we should talk. I'm also somewhat interested in the crowdfunding aspect and technology behind it.
Yeah I came across that a few weeks ago. Glad to know you are making progress. Can you contact me via hello@storj.io, and lets figure out some ways we can work directly together. Storj doesn't believe in competition, we instead want to find ways to collaborate with those in the decentralized data space.
Why does storj need 6 million dollars? What are they planning on doing with it?
One problem with Storj is that their technical contributions to the field have been very minimal, the best I can tell is that they've introduced 'BitCumulis', which is a 'modular' design for integrating multiple solutions such as those presented in MaidSAFE and ethereum. They haven't actually made any convincing contributions to the bigger problems.
From what I can tell, storj is a front-end that is waiting for other teams to solve their proof of storage and proof of bandwidth problems. And I don't understand how that calls for 9800 BTC.
For those who don't know about the MaidSAFE project,
The Whitepaper (a good read): https://github.com/maidsafe/Whitepapers/blob/master/Project-...
here's their code: https://github.com/maidsafe
This project is something that's quite interesting in its scope. It goes to see how successful they'll be in implementing all of this, and how robust the system would turn out (e.g. would someone lose data if 10% of the network magically disappears overnight)
IMO this deserves a lot more attention than storj. It's everything storj has, plus a whole decentralized internet infrastructure upon which other decentralized applications can be built. It seems the most logical way forward.
So Metadisk is out, and DriveMiner is on its way. As far as technical contributions, going to have to wait until I finish up the revised whitepaper.
We were focused on writing the code first, and now we are in the process of writing down what we learned. Don't take me at my word, but do give me a few weeks to prepare a more formal paper now that most of our core code has been implemented or speced.
As far as the $6M we don't intend to raise that much. What is left over will be distributed through the early mining, and future crowdsales for other platforms build on top of Storj(decentralized Youtube, Imgur, etc.) In that way we can give the network a kickstart with the actual resources that it needs to survive.
I wish there was some way to participate in this using my NAS. I don't want to keep a PC online all the time just to service other folks' storage requests.
Sounds like a neat idea for a homebrew Synology package.
They need to store things in the blockchain, namely meta-data. I think there are also a few other changes that they made to the blockchain structure that would have made it very hard to simply use Bitcoin.
I imagine that they'll at least merge-mine with Bitcoin.
They don't want to host storage space. They want to bootstrap a cryptocurrency backed by storage space and availability as value. Eventually it would lead to a decentralized storage, and you would pay with the cryptocurrency for it.
23 comments
[ 5.7 ms ] story [ 57.3 ms ] threadWhere can I find some precise protocol specifications? The whitepaper is full of handwaving about the critical parts (namely Proof of Resource).
They (last I checked) are approaching the problem by using quadruple redundancy, no erasure coding. This opens some attack vectors, because an attacker only needs to control 4/4 pieces, something that takes a probability of (attacker control)^4 . Their proof of storage also seems to use nonce's, which require precomputation before uploading the file, and means a limited amount of proofs before you have to fetch the file and compute a bunch of other nonce's.
The biggest problem with MaidSAFE's stuff is that it's all over the place. There's no concise document that explains proof-of-storage. Their approach seems enormously complex, which to me indicates a huge likelihood of undetected vulnerabilities (complexity is bad when you want security).
That paper was written before decentralized storage was really just a new idea. I've since implemented a baseline core algorithm, and we have one application running.
I'm in the process of rewriting the whitepaper. Hopefully will put out a draft in two weeks(I'll be at conferences all this week). From there I want to actually collect network data(after all I'm trained as a researcher and what is good paper without any data?) before putting out a solid draft to be peer reviewed.
http://filecoin.io/ http://filecoin.io/filecoin.pdf
Also I think filecoin has a mistake. If it ever gets popular then it is possible that the difficulty would be quite high (in the scale of GBs). In that case you can't effectively mine only if you share at least 1 TB of your hard drive below that the expected reward for a block would be near 0. It would drive the network to centralization. The expectation reward for a block isn't linearly proportional to your shared disk space but a convex function of it.
On the other hand it would be possible to implement mining pools which could somewhat solve this problem.
I'm at a slightly different layer of the stack - one that might be advantageous for users wanting to run decentralized offerings like Storj and Maidsafe without having to do the deployments themselves. If any of the Storj guys are interested, we should talk. I'm also somewhat interested in the crowdfunding aspect and technology behind it.
One problem with Storj is that their technical contributions to the field have been very minimal, the best I can tell is that they've introduced 'BitCumulis', which is a 'modular' design for integrating multiple solutions such as those presented in MaidSAFE and ethereum. They haven't actually made any convincing contributions to the bigger problems.
From what I can tell, storj is a front-end that is waiting for other teams to solve their proof of storage and proof of bandwidth problems. And I don't understand how that calls for 9800 BTC.
We were focused on writing the code first, and now we are in the process of writing down what we learned. Don't take me at my word, but do give me a few weeks to prepare a more formal paper now that most of our core code has been implemented or speced.
As far as the $6M we don't intend to raise that much. What is left over will be distributed through the early mining, and future crowdsales for other platforms build on top of Storj(decentralized Youtube, Imgur, etc.) In that way we can give the network a kickstart with the actual resources that it needs to survive.
Sounds like a neat idea for a homebrew Synology package.
I imagine that they'll at least merge-mine with Bitcoin.
If sidechains or treechains become viable we plan to merge it back in to Bitcoin proper.