Hacker News Reader

Comment by Taek | Hacker News Reader

Maybe I've missed something, but how does Filecoin handle redundancy, and how does Filecoin handle delegation?

The devs seem to hope that by making the rare pieces valuable, they will somehow avoid having low redundancy. But what happens when everyone joins a large pool like Ghash.io, and every file has only a single point of failure?

I would not trust my data to Filecoin, it seems too easy for files to get permanently lost. There are no guarantees that every file piece will be managed by at least N nodes.

View on HN

Great questions :) -- this borders into future announcements, so sadly I'll have to defer to the answer on this paper, which you summarize well here. (Not asking you to believe, only to wait a bit longer.) Also, this answer is actually not bad: information on the distribution of pieces will become widely available so users can tune rewards accordingly. Storage pools also have a ceiling: after backing up the entire piece set, it's better for them to compete than to keep growing.

After backing up the entire piece set, the requirement per-node for joining the pool declines - say 100GB each instead of 500GB each. Or perhaps you have a setup where the pool master takes care of storing all the data and the workers all do POW. Then it's up to the pool to tune prices so that workers can profit as much as possible.

You might end up with multiple competing pools, but it puts the barrier-to-entry of being a new pool at a high price. Everyone will mine for the cheapest pool, and the 1 pool that can provide all the storage at the lowest cost is the 1 pool that will control all of the data on the network.

It seems like a fragile solution.

In equilibrium, people will still compete to avoid sharing the reward with others if they can avoid it. You don't need a full pool to have high expected reward. Anyway, stay tuned ;) (+ for getting rid of PoW altogether).

Maybe you could also expand to also allow poeple to mine by running a distributed search engine. So they can basicly decide wether they offer simple storage or storage + processing power + low lattency.

Store redundant copies yourself, each copy encrypted with a different key.

If there are a lot of storage nodes competing on price each copy should end up on a different node, and none would be able to afford a centralized storage service.

Reply on news.ycombinator.com