5 ms·

[Author here] short answer taken from our .PDF file: "Considering that Dropbox [53] currently holds around 500 petabytes of user data [54], one could argue tha

[Author here] short answer taken from our .PDF file: "Considering that Dropbox [53] currently holds around 500 petabytes of user data [54], one could argue that Filecoin is overvalued." Study we conducted with a master student at Delft University of Technology. Open lab notes when writing this paper: https://github.com/Tribler/tribler/issues/3097 https://github.com/Tribler/tribler/issues/3097 Note TUDelft has currently 8 professors in their http://blockchain-lab.org http://blockchain-lab.org Various scientists and 38 master students working on improving our own Filecoin-like system, based on our Trustchain fabric: https://github.com/Tribler/tribler/issues?q=is%3Aissue+is%3Aopen+label%3A%22MSc+course+work%22 https://github.com/Tribler/tribler/issues?q=is%3Aissue+is%3A... It's home to one of the largest blockchain labs of Europe.

eco 9y ago
Dropbox feels like the wrong thing to compare it to. IPFS isn't really for storing files privately. Public S3 buckets would be closer. I don't have any numbers to say if it's overvalued based on how much public cloud storage there is (I wouldn't be surprised if it were though).
- api 9y ago
  Storage costs are hyper-deflationary. We have a NAS on a desk with 27 terabytes of storage and it cost under $1000 (with drives) and consumes about 40 watts of power. Ten years ago this would have cost closer to $5000-$10000 and consumed hundreds of watts of power. We bought it about a year and a half ago and already we could replace it for less than $500. The idea of distributed storage at the edge is interesting but it's not something that would cost $250 million to build. We could probably build a friendly system for doing this for under $500k in developer time. It's also a mystery how efforts like MaidSafe or whatever it's called this week have never managed to ship anything after getting multiple multi-million-dollar golden showers. Distributed storage of immutable content-addressable blobs is not that hard of a problem. BitTorrent kind of solved it long ago, albeit with a different UI/UX metaphor and use case. Just take those ideas and pivot them a little and wrap them around the S3 UI/UX metaphor and add payment channels or some other accounting mechanism and you're done. You don't necessarily need the sexiest tech. In fact sexy tech probably makes it more fragile and harder to use. DHT + hashing + RAID-type redundancy techniques + wallets will do it. Everyone over-thinks and over-engineers things these days and it drives me crazy. Stop it. Intelligence is for the things you can't do, not for doing the things you can already do in more baroque ways. Intelligence is also for simplification. Complexification is stupidity.
  - jacquesm 9y ago
    Sorry to see this downvoted, if this were reddit there'd be some reddit gold right here.
    - synctext 9y ago
      > never managed to ship anything after getting multiple multi-million-dollar golden showers. It's more then just linking hard-disks. It is really hard to build the incentives and market part. Read for instance about our operational prototype from 2014, "Towards a Peer-to-Peer Bandwidth Marketplace", https://link.springer.com/chapter/10.1007/978-3-642-45249-9_20 https://link.springer.com/chapter/10.1007/978-3-642-45249-9_...
      - api 9y ago
        So ship the first thing then do the second thing. You get two things from this. First, your storage system gets battle tested. Second, you get the credibility of joining the ranks of those who can actually ship quality software.
  - ethanwillis 9y ago
    Siacoin with minio integration(fully integrated as of last week) allows you to have S3 compatible "edge storage" right now.
  - wmf 9y ago
    There's a lot to agree with and disagree with here. Sia is currently (barely) working and was built for $1.25M, so $250M clearly isn't needed (except maybe for marketing, which generally costs more than development). I don't think the approach of "the simplest thing that could possibly work" applies to the decentralized/blockchain space. When money is at stake, attackers come out of the woodwork.
  - kodablah 9y ago
    > Distributed storage of immutable content-addressable blobs is not that hard of a problem. To do so equitably without the ability to be gamed is quite difficult. BitTorrent is quite game-able so we end up with closed communities and private trackers. To handwave these adversarial concerns away is unhelpful. Implement a s/kademlia DHT with all of the security conventions in latter papers and it becomes clear how hard it is to do safely. I am confident IPFS's DHT could be quite easily harmed (most are implemented around hash bucket closeness) if it were valuable enough to do so.

eco 9y ago

Dropbox feels like the wrong thing to compare it to. IPFS isn't really for storing files privately. Public S3 buckets would be closer. I don't have any numbers to say if it's overvalued based on how much public cloud storage there is (I wouldn't be surprised if it were though).

api 9y ago
Storage costs are hyper-deflationary. We have a NAS on a desk with 27 terabytes of storage and it cost under $1000 (with drives) and consumes about 40 watts of power. Ten years ago this would have cost closer to $5000-$10000 and consumed hundreds of watts of power. We bought it about a year and a half ago and already we could replace it for less than $500. The idea of distributed storage at the edge is interesting but it's not something that would cost $250 million to build. We could probably build a friendly system for doing this for under $500k in developer time. It's also a mystery how efforts like MaidSafe or whatever it's called this week have never managed to ship anything after getting multiple multi-million-dollar golden showers. Distributed storage of immutable content-addressable blobs is not that hard of a problem. BitTorrent kind of solved it long ago, albeit with a different UI/UX metaphor and use case. Just take those ideas and pivot them a little and wrap them around the S3 UI/UX metaphor and add payment channels or some other accounting mechanism and you're done. You don't necessarily need the sexiest tech. In fact sexy tech probably makes it more fragile and harder to use. DHT + hashing + RAID-type redundancy techniques + wallets will do it. Everyone over-thinks and over-engineers things these days and it drives me crazy. Stop it. Intelligence is for the things you can't do, not for doing the things you can already do in more baroque ways. Intelligence is also for simplification. Complexification is stupidity.
- jacquesm 9y ago
  Sorry to see this downvoted, if this were reddit there'd be some reddit gold right here.
  - synctext 9y ago
    > never managed to ship anything after getting multiple multi-million-dollar golden showers. It's more then just linking hard-disks. It is really hard to build the incentives and market part. Read for instance about our operational prototype from 2014, "Towards a Peer-to-Peer Bandwidth Marketplace", https://link.springer.com/chapter/10.1007/978-3-642-45249-9_20 https://link.springer.com/chapter/10.1007/978-3-642-45249-9_...
    - api 9y ago
      So ship the first thing then do the second thing. You get two things from this. First, your storage system gets battle tested. Second, you get the credibility of joining the ranks of those who can actually ship quality software.
- ethanwillis 9y ago
  Siacoin with minio integration(fully integrated as of last week) allows you to have S3 compatible "edge storage" right now.
- wmf 9y ago
  There's a lot to agree with and disagree with here. Sia is currently (barely) working and was built for $1.25M, so $250M clearly isn't needed (except maybe for marketing, which generally costs more than development). I don't think the approach of "the simplest thing that could possibly work" applies to the decentralized/blockchain space. When money is at stake, attackers come out of the woodwork.
- kodablah 9y ago
  > Distributed storage of immutable content-addressable blobs is not that hard of a problem. To do so equitably without the ability to be gamed is quite difficult. BitTorrent is quite game-able so we end up with closed communities and private trackers. To handwave these adversarial concerns away is unhelpful. Implement a s/kademlia DHT with all of the security conventions in latter papers and it becomes clear how hard it is to do safely. I am confident IPFS's DHT could be quite easily harmed (most are implemented around hash bucket closeness) if it were valuable enough to do so.