Quote from: TPTB_need_war on February 08, 2016, 12:18:50 PM

Decentralized proof-of-storage is always flawed, and that will include IPFS's Filecoin as well.

In addition to the insoluble issue of how to not infringe copyright in decentralized file storage (and thus per my prior three posts above not be at odds with the coming Knowledge Age creators and the government), another point is how are we going to pay for decentralized file storage for all websites and content on the internet Huh

If we say that we'll all host this content for free on our home computers and serve it via our home ISP internet connections, I have already pointed out insoluble problems:

Last mile physics is such that home connections have asymmetric upload (less than download) bandwidth. Thus even if we all trade our bandwidth tit-for-tat equally, then we have collectively more download bandwidth than we have upload bandwidth. Not only does this steal from ISPs that provide more upload bandwidth, which is what I warning Bittorrent in 2008 would lead to the Communism of Net Neutrality (and they ignored me and just this month they ostensibly removed the archives when I raised this point again!), but it also is impossible to balance and thus we pay for it by hosting it on symmetrically backbone connected hosts as the internet is currently. So then how do we pay for that upload and hosting service if the owner of the files is not hosting them? Do we send microtransactions for each file we request, i.e. we put a paywall on the entire internet(!!!)? Or do we pay a Communist tax to the government (e.g. the plans for Net Neutrality) and it regulates a file store for us and also so the government can enforce copyright? I guess one can argue that bandwidth will become so cheap, we can afford to consider it free. But still someone has to pay for it and even as costs decline, bandwidth demands increase. And global bandwidth costs are I would guess in the $billions.
There is no way to insure persistence unless one keeps a copy of the file hosted themselves, thus this sort of defeats one of the main point of decentralized file storage which is reliability of archival.

I do think we need smarter decentralized caching for the internet which pays for itself by saving bandwidth! I do think we need to reference files by hash (perhaps with a hint for primary source url where also blame can be assigned for promulgating copyright infringement, thus all such records would need to be signed with the public key for the HTTPS certificate of the site) for permanence of references.

So perhaps there is a way to do a decentralized overlay network on top of the URL paradigm that can deal with copyright blame and respect the policies so encoded by the aforementioned signature. I am going to email Juan Benet again and pass this idea along. I am urging him and others to be less ideological and more pragmatic.

Delusional leanings such as proof-of-storage (which I thought of in 2013 and discarded because it can't be sound) and wanting to attack the establishment at any cost of nonsensical economic, political-economic, and technological issues. Let's get back to standing on solid engineering work.

Quote from: TPTB_need_war on February 08, 2016, 07:41:45 PM

Quote from: super3 on February 08, 2016, 05:01:48 PM

Quote from: TPTB_need_war on February 05, 2016, 07:20:33 PM

All decentralized file systems are doomed:

https://bitcointalk.org/index.php?topic=1344997.msg13730325#msg13730325
https://bitcointalk.org/index.php?topic=1340057.msg13670558#msg13670558
https://bitcointalk.org/index.php?topic=1340057.msg13783053#msg13783053

Storj guy here. All of the actor vectors you described are addressed in the our whitepaper or a blog post.

I read the white paper last week.

Quote from: super3 on February 08, 2016, 05:01:48 PM

1) Sybil - Bonds and unique pieces

There is no way to do proof-of-storage that is robust. The only way is to make some assumptions about latency of propagation to a centralized copy of all files, but that can be gamed. Propagation is not proof.

Quote from: super3 on February 08, 2016, 05:01:48 PM

2) Illegal content - Greylists

The Storj FAQ confirms these are opt-in, and not forced. Thus I maintain my point that the Storj protocol can become banned (refused) by Hosts (and even ISPs). We are moving into totalitarianism and increased government control over the internet.

This direction of enabling theft of copyrights is begging for your project to be attacked and fail.

Quote from: super3 on February 08, 2016, 05:01:48 PM

3) Bandwidth vs storage - pay for both

Pay how? Micropayments for each access to bandwidth?

How to pay for storage when it is decentralized with unbounded replications and can be Sybil attacked.

Sorry these decentralized systems are doomed. The concept can't work.

Quote from: TPTB_need_war on February 10, 2016, 04:59:26 AM

Quote from: super3 on February 08, 2016, 10:09:08 PM

Quote from: TPTB_need_war on February 08, 2016, 07:41:45 PM

Quote from: super3 on February 08, 2016, 05:01:48 PM

1) Sybil - Bonds and unique pieces

The mathematical models are right there in the paper. We are collecting live data from the network, which proves the models are correct.

"Its not going to work" in face of real data showing that is working is not going to cut it. Please provide some data or mathematical models that say otherwise. Latency doesn't matter for proofs.

Testnets do not prove that the Sybil attack resistance and payment model economics work (because game theory is fully incentivized in the wild).

Regarding case 1) in the quote above, the fact is the math models are often myopic[1] (and again that is so in this case), because it is impossible to prove proof-of-storage/retrievability:

These proof-of-storage/retrievability algorithms also employ a challenge/response to force the node to have access to the full copy of the data which should be stored, but this does not prevent the node from outsourcing the storage to a single centralized repository. So to attempt prevent that centralized repository attack (i.e. Sybil attack on the nodes) these proof-of-storage/retrievability algorithms try to use network latency to prevent centralized outsourcing, but [that is impossible because] ubiquitously consistent network latency is not a reliable commodity.

[1]

Meni Rosenfeld's myopic math, and note Meni is a widely respected academic:
https://bitcointalk.org/index.php?topic=1319681.msg13633504#msg13633504

And my explanation of the myopia:
https://bitcointalk.org/index.php?topic=1319681.msg13797768#msg13797768
https://bitcointalk.org/index.php?topic=1319681.msg13819991#msg13819991
https://bitcointalk.org/index.php?topic=1319681.msg13763395#msg13763395
https://bitcointalk.org/index.php?topic=1319681.msg13647887#msg13647887

Quote from: ArticMine on February 09, 2016, 04:03:16 AM

Quote from: TPTB_need_war on February 05, 2016, 07:20:33 PM

All decentralized file systems are doomed:

https://bitcointalk.org/index.php?topic=1344997.msg13730325#msg13730325
https://bitcointalk.org/index.php?topic=1340057.msg13670558#msg13670558
https://bitcointalk.org/index.php?topic=1340057.msg13783053#msg13783053

Of course decentralized file systems are not compatible with the 19th century business models of the MPAA and RIAA, such as placing music on Edison Cylinders https://en.wikipedia.org/wiki/Phonograph_cylinder and then distributing the Edison Cylinders by sailing ship. That does not mean however that decentralized file systems are doomed, what it does mean is that decentralized file systems will serve to further accelerate the demise of these 19th century business models, and the corporations that promote them.

The political argument is irrelevant for as long as the technology and payment (economic) models are irreparably flawed, as I explained above.