You are right. The paper should say "connected to at least 60%", so the nodes should be connected to more than 60%, maybe 70% or more to prevent the problem you said.
The objective of my idea is not to have cloud computing p2p, but to make better use of the cpu/gpu power. I like the idea of p2p currency, a lot indeed, but I also want to make a better use of the energy consumed, that's the reason I'm proposing the cloud service, to make useful calculations that while secure the net and the currency, also provide solutions to other problems and maybe profit from a fee, adding value to the currency.
With a currency that requires to be able different calculations and functions like this, it would be almost impossible that someone makes a custom device that broke the system and wins the 51%, because one don't know what is required to solve the next problem, maybe it is simply a hash, maybe it is a collection of Fast Fourier transformade, maybe it is a block of data to be transferred, maybe it is a full simulation of a protein molecule, who knows? and that is going to be impossible to predict and impossible to make a custom device.
About the cost for cloud computing, what you are getting is your block solved with your own kernel code, in at least 60% of the total nodes of the network, so if there are 100 nodes, you know that at least 60 nodes solved the problem with equal results. If something is wrong in the result is your fault, because you provided a bad kernel to solve it, and you need to fix your code and resubmit the block and the new kernel to solve it.
For storage, the idea is that in some countries the public TV is a garbage, with manipulated media, so an open TV, with people making it's own media, that also provides the big movie producers a way to have DRM or charge for the addition of video, that can't be handled by any entity, because it is open and p2p like namecoin and bitcoin, could replace the public, manipulated TV. That will require a custom device that can be made with very cheap hardware (about 35-50 USD, check Raspberry PI or another micro computers), a GUI to present the video and menus to the viewer and the cloud storage.
One can setup that micro computer box, share the space of the flash card and the bandwidth to the network, win some coins while doing that, and use that coins to be able to consume blocks from the network and also to pay for DRM protected media. That payment is going to be redirected to the owner of the block of data, with a commission of the miners that relayed that block, and also that could add proof of work to the coin, because the block is required to be relayed by 3 nodes, generating work on that 3 nodes that require to confirm that the block was relayed by them.
cloud coin is still a simple, napkin idea hehehe
You do not seem to address the rather important matter of consensus.
If my node thinks 60% of everyone's data has arrived and thus closes the block, what happens with all the other versions of the block who heard from a slightly different 60% than I did? Each of us when we were at 59.9999% might not have had the same nodes in our collection but even if we did, we each could hear from a different node as our last straw that broke the 60% barrier and caused us to close the block.
Also, since your objective supposedly is to be a provider of cloud computing, all of this crap about how currencies could work is irrelevant, as there are already working currencies in existence. What you need to do as a provider of distributed p2p cloud computing is solve that goal's problems, such as how to decide which nodes to pay how much currency to and why.
As a person wanting cloud computing, how do you check you got what you wanted and are not being charged for services that were not in fact provided, stuff like that. See for example the distributed storage thread about dirt cheap remote disk storage which so far it seems turns out not to be dirt cheap at all due to being a difficult problem; notice how often they end up back at the idea of a centralised service that will co-ordinate home computer nodes that want to be paid for providing storage... And to cloud-compute, you need storage out there to hold data people want computation to be done with/on so storage is just part of the problem you are facing...
-MarkM-
