The network viewing a LAN or cluster of containers as a single entity, for the sake of data integrity, won’t stop a university dorms spare resource being part of the network. That LAN is still entitled to store 1/6th of all data on the Safe Network, which would be thousands, if not millions of times larger than the LAN itself. And if the network is designed properly, they should be rewarded as if they were individual machines.
We could discuss this forever, but we will only know the more detailed aspects/behaviour of the network once we have a more complete version. Once I get my hands on that I hope to enhance behaviour (if it doesn’t already exist) in a few areas.
Node ageing will consume offered space on the node at a set rate through disposable data. This means the network can discover nodes trying to game the system using PoR before it actually needs the resource.
Cluster the the entire Safe Network into 6 groups, based on a simple “ping” between nodes to establish a rough distance. Each copy of a chunk will be distributed between these groups. This means the network has global data integrity, and will reduce latency, as at least one copy of the data will be in the cluster you are part of, and 2 more near you.
I suspect it may be easier to assert/encourage that everyone will run lots of vaults to ensure an even distribution.
Slow vaults will get few rewards and will potentially lose age over time. Rather than try to dictate network topology, ensuring abuse isn’t rewarded should be the goal.
It’s not the quantity of nodes that determines the success of the safe network, it’s the quality.
I get lots of people want farming to be super easy and profitable by throwing up 1000s of containers backed by virtual volumes of cheap cloud storage (such as the unlimited google drive), to make as much money as possible.
MaidSafes biggest barrier to success is to get everyone to trust that the network is the safest place to store valuable data.
Some Dude: How is data store of the safe network to keep it secure and safe.
MaidSafe: We encrypt your data and store 6 copies on random nodes over the network.
Some Dude: So copies could all end up in the same university dorm?
MaidSafe: Yes, but it’s incredibly unlikely.
Nobody cares about the but, they will simply focus on the fact it’s possible. The network seems to provide no guarantees of data safety. Sometimes in software you need to implement things because it’s not backed by math, or science, but simply for humans. Look at Apple and how successful they became in the early days by focusing on the “Human” interaction with technology.
Edit:
Also, if someone forked the SafeNetwork to add those two guarantees in my previous post, even if they achieve nothing, and provide no extra data redundancy 99.99999999% of the time or protection to the network. Which network do you think people will pick? Probably the one that markets itself as providing data guarantees.
They are not super complicated “features”, and I’m not saying we need them for beta. However, in terms of fuelling adoption, I think they are important to consider.
It is possible to create satoshi’s keypair on bitcoin as well. Possible in any decentralised network is a mathematical possibility, probability is very important and computationally/problematically irrelevant is an important concept. Given enough nodes the network is not going to store all data in a uni dorm room and that statement is with probability near 1.
I would like to show you how it should work, but I can not find it. But to secure each copy of chunk, there is several hops between each copy. So the university would need several different providers to be able to store all copies in one area. But than malfunction of one provider would not effect the others. It is also very small chance that all copies will be in a same country.
Just wondering here: Would it be desirable/feasible to program the network to reject a hosting node if that node already has one chunk of the same file? The chunk would then continue its journey to the next random location. Would this create too much of a drag on the flow of data? Maybe this feature could only be enabled during the early stages of the network’s development, when the chances of one node having multiple chunks of the same file are the greatest.
No, there is no link at all between chunks. This itself is a security measure, where no node knows it’s chunks are related to any other. Only public data could do this if the data map older could read the data map (it can), but then it has no authority to tell other nodes anything about accept/reject store locations.
To help here (this will sound cheeky, it is not supposed to be), we select immutableData store location based on the SHA hash of the data. the security of SHA comes from hashing any data will equally distribute across an address range (2^256 in our case). The numbers are amazingly high and the chance of collision thought to be infeasible. I really don’t see an issue here, it would be like us trying to stipulate the private keys people could have. It is infeasible and also means complexity like you would not believe for a problem you cannot feasibly have. That complexity is almost infinite as you would need to almost build another universe or a secure hash mechanism where the keys were larger than the square root of the data length + data length(). That would be insane though and you are better using encryption, then you get back to what if I encrypt 2 pieces of data and the CipherTex has similar output, then it could end up in the same dorm room of a uni … and on you go
As I read it @zeroflaw is bringing up the possibility that all copies of a single chunk could end up residing on one physical machine or domain, rather than all the chunks that make up a file ending up there which is much less likely.
So therefore if that machine or domain were lost the chunk would be lost too and the file it is part of would be unrecoverable. At least while the network is small, this is a possibility. The probabilty is low, yes, but maybe significant while the network is small and before it becomes truly global. But compared to other problems of a small network (Sybil attacks etc) I’m guessing it’s a minor issue, but it’s one more reason to make sure the network grows as quickly as possible.
Nothing is 100% secure (can we rule out an astroid strike, a zombie apolcalypse?) but if we can say it is X times more secure and available than AWS and have numbers to prove it that will be a very good marketing message.
And that is why David replied to explain why that is not such a problem. Also the issue he has is that he thinks that the number of nodes is not important but the quality of the nodes is.
But to solve the issue that you highlight @JPL the number of nodes is very important. The quality will be controlled by the network itself as inferior ones will be weeded out. This is why i replied previous to say his analysis is not good. Its spinning a narrative rather than exploring the issues
I built and ran SNAPP recently and haven’t seen much change since so am not expecting it for a bit yet. I tweeted a screenshot - it looks very nice. You fellas should follow Twitter.com/safepress!
