Quote from: iamnotback on June 27, 2016, 03:18:59 PM

Quote from: super3 on June 27, 2016, 02:48:21 PM

Quote from: iamnotback on June 22, 2016, 06:42:49 PM

Quote from: super3 on June 22, 2016, 06:35:01 PM

Quote from: iamnotback on June 22, 2016, 05:59:23 PM

Quote from: super3 on June 22, 2016, 05:16:51 PM

Wait a minute, how did you handle redundancy in your old solution? Did you do something like 3x redundancy in case some of the nodes went down?

I didn't develop (pursue) the idea beyond the conceptual investigation phase, because I determined that it wasn't a solid enough direction to pursue.

The idealism of it appeals to me of course. But I've also learned to be very skeptical of idealistic causes, because they can be intoxicating and cloud objectivity.

I am obviously going to be more circumspect about dubious project technologies, given my age. I don't have another decade to expend on something that does not pan out.

Everything starts as an idea. Do you believe in the idea of distribution and decentralization? It all falls apart if we can't get our data out of centralized data centers. What good is a decentralized application if its just run at Amazon S3?

Of course I do.

I'll paradigm-shift you. We can decentralize our servers. Abstractly I am thinking the fundamental error in decentralized file stores such as these, is we are modelling a monolith, i.e. a total order on redundancy. Paradigm-shift to a plurality of partial orders.

Btw, I like the name Storj.

Thanks its taken with permission from this post: https://bitcointalk.org/index.php?topic=53855.msg642768#msg642768

Kudos to Gmaxwell on the name then.

I have an idea.

What about a different approach to achieving redundancy.

Redundancy is fundamentally about making sure our data is stored on more than one hard disk.

If we could disperse the bits of the data across TBs of data, then the host actually has no incentive to cheat as the host can use RAID striping to maximize their performance.

So then we probably need a blockchain to manage this coordination.

Best way to do this is through Reed-Solomon erasure encoding. Its kinda like a mini RAID per file. For example we use 20-of-40. So the file is broken into 40 pieces of which you need any 20 to recover. Like you said you can't solve Sybils, but proper erasure encoding can make it many orders of magnitude more resistant.