I did not understand the explanation, why is the data statistically likely to be stored on multiple HDs (in distinct locations)?
@super3 claimed a hyperbolic distribution. Imagine all the hard disk as holes on a table and you toss a coin and it randomly lands in any hole (surely you've played this game where even if you aim for a hole, you don't always get it). Disks only have finite storage so one disk can't be a hole the size of the entire table.
But yeah, it is not going to end up being dominated by home computers on consumer Internet connections. In addition to the economic (economy-of-scale) reasons the server farms will dominate, users of cloud storage also don't want their redundancy to be dependent on unprofessionally, (un)managed, home computers that can go down for numerous reasons (no backup power, no RAID, thief breaks into home, virus attack, etc).
I do not see the issue for the latter point. Presumably it could be determined that x unprofessional hard drives are equivalent to y server farms, and an algorithm could take this into account in determining number of copies of data to distribute where.
But you don't know what quality of unprofessional you have. The lowest quality might be 1000X less reliable that the best. Yet the price might not be a even 100X cheaper.
And my main point is I don't see the point of the tokens. I don't see any investment value of investing in the tokens long-term. They may get pumped by hype but in the long-run we will be using "Bitcoin 2.0" (what ever that ends up being) to pay for these services, not Sia, Storj, MaidSafe tokens.
I never saw any value in the tokens after I investigated how they worked. But the cost to store data at both Storj and Sia are much cheaper than the current market offerings (you pointed out some reasons why). I need another copy of some data, I might use one or the other for storage.
Anyone who has run a business successfully understands that 10X cheaper runs the risk of 1000X more losses. Hire the best, produce high quality, prosper.
You think 99% uptime is okay until you missed on a $1 million deal because the offer was only available during the 1% downtime.
Edit: more on that here:
In addition to
my prior critique of "proof-of-storage", I see some additional flaws in the idea expressed as quoted below:
Coins are issued by the network based on the following formula:
-1 coin = 1gb hosted for 1 month.
-Any downtime (detected by pinging) reduces profit 10x (ie, if your mining machine is down for 1 day, you lose 10 days worth of profit for that uptime month)
-100% of your "storage" has to be downloadable by the network within 1 hour, tested by the network randomly 4 times per month (uptime month of 30 days, not calendar). If you fail this test your profits over this period are reduced by double the amount of failed download, eg, you are hosting (mining with) 4gb of space, a random download attempt occurs and only 90% of the 4gb is downloaded, then your profits are reduced by 20% untill the next random download test.
-When you start mining you do not receive profit for the first uptime week of 7 days (this is to stop people that had some downtime simply creating a new miner on a new wallet straight away)
- Ping checks are performed every 15 minutes, you need to fail 2 to be considered "down". Thus you can install an update and restart without "down time".
- Miners are also rewarded the transaction fees of the network, spread evenly to the miners based on earning.
None of these quoted are
objectively provable to the public-at-large, i.e. on a blockchain. For example,
proof-of-storage can work from the perspective of the owner of the data to be stored, but not from a public perspective.
Network performance can't be proven. This is one of the fundamental reasons we have to deal with Byzantine fault tolerance on networks. How do you prove to a blockchain that the ping time you measured was accurate. You can't. How do you prove downtime. You can't. If you say voting, then you have Sybil attacks on voting. Byzantine agreement can't remain unstuck without a hardfork or whales. Etc..
Sorry this is entirely impossible. It violates the basic research and fundamentals. Much more detail is in my unpublished white paper wherein I start from first principles and try to explain these fundamentals (but ends up being far too much to summarize to laymen, so I don't know if that version of the whitepaper will be the one I end up publishing).
So this is what I mean with my criticism that proof-of-storage can't even really work well even for file storage in the Storj model where each user encrypts the data to be stored (in multiple variants), because it is impossible to insure fungible performance for the data retrievability.
