8

I have talked with pembo a little bit about possibly making Phuks into a decentralised p2p type site. Similar things already exist. Aether is one that has been around for a while now, I remember trying to connect to it during the Reddit blackout and I couldn't connect, probably because they were having the same mass influx problems as Voat. I recently came across another called Zeronet, which is a little different due to it being a kind of network of sites within the same client, so it isn't just an aggregator, it has other parts too. It doesn't have that much content at the moment though.

I have also been reading about other kinds of decentralised networks like pirate boxes, and something called the Briar Project, which is a mobile meshnet kind of thing that is (apparently) used by journalists and activists during internet blackouts/for secure communication. I don't fully understand how it works, but it essentially creates a huge LAN between phones, and it also uses bluetooth in some way.

My question is this: Will some form of this technology eventually replace ISPs?

If you were to take a densely populated developed area like London, for example, would it be viable to create a LAN for the entire city? It seems to me like it could work with current technology. There are mobile devices metres away from each other, without even taking computers into account. If it was possible to do this, what would it take to get people to make it happen? It seems like not having to pay an ISP would be a major incentive, but there is also the privacy aspect too. Over the next few years, laws relating to the internet are probably going to become more and more invasive. This has basically already happened in the UK.

Could the internet of things actually serve a purpose in a project like this?

There are cameras, fridges, toasters and all sorts of ridiculous things that are connected to the internet now. If a lot of appliances had the capacity to join/support a network like this and the range was improved, would it be possible to have a decentralised world wide LAN? Is it even a good idea, considering the security risks that it could produce?

It seems to me like it could be part of a natural progression that is happening. We ditched wired connections for wifi, and now there is potential for something else.

I have talked with pembo a little bit about possibly making Phuks into a decentralised p2p type site. Similar things already exist. Aether is one that has been around for a while now, I remember trying to connect to it during the Reddit blackout and I couldn't connect, probably because they were having the same mass influx problems as Voat. I recently came across another called Zeronet, which is a little different due to it being a kind of network of sites within the same client, so it isn't just an aggregator, it has other parts too. It doesn't have that much content at the moment though. I have also been reading about other kinds of decentralised networks like pirate boxes, and something called the Briar Project, which is a mobile meshnet kind of thing that is (apparently) used by journalists and activists during internet blackouts/for secure communication. I don't fully understand how it works, but it essentially creates a huge LAN between phones, and it also uses bluetooth in some way. My question is this: Will some form of this technology eventually replace ISPs? If you were to take a densely populated developed area like London, for example, would it be viable to create a LAN for the entire city? It seems to me like it could work with current technology. There are mobile devices metres away from each other, without even taking computers into account. If it was possible to do this, what would it take to get people to make it happen? It seems like not having to pay an ISP would be a major incentive, but there is also the privacy aspect too. Over the next few years, laws relating to the internet are probably going to become more and more invasive. This has basically already happened in the UK. Could the internet of things actually serve a purpose in a project like this? There are cameras, fridges, toasters and all sorts of ridiculous things that are connected to the internet now. If a lot of appliances had the capacity to join/support a network like this and the range was improved, would it be possible to have a decentralised world wide LAN? Is it even a good idea, considering the security risks that it could produce? It seems to me like it could be part of a natural progression that is happening. We ditched wired connections for wifi, and now there is potential for something else.

6 comments

[–] phoxy 3 points (+3|-0) Edited

I hate to say it but Voat and Phuks are temporary for me. I'm waiting for a solid secure distributed p2p discussion platform to rise. Aether looks promising but I have no information on its security or anonymity. And it needs a web based UI. And I don't know if it has solved the problem of a malicious entity with many identities influencing the discussion (disproportionate ratings, not shilling)..

[–] PMYA [OP] 3 points (+3|-0)

And I don't know if it has solved the problem of a malicious entity with many identities influencing the discussion

This is unavoidable on any site with profiles I think. At least, it is unavoidable without either stripping down voting/sorting features, or implementing things that infringe on user privacy.

Polsaker found a way to possibly stop or severely restrict vote brigading, but it would raise privacy concerns. A combination of locking votes to IPs and browser fingerprinting could stop the brigading issue, but people might not be happy about the site storing their browser information.

[–] phoxy 0 points (+0|-0)

There are algorithms (distributed or based on distributable calculations) which can detect abnormalities in a network of peers. Graph theory has a number of different centrality/influence/power measures which rank the vertices of a graph based on the structure of the graph. Google uses one of these centrality measures to determine page importance and it's what made their results so much better than their early competitors.

Another measure of centrality that I find interesting is called PN centrality (PDF) (positive/negative) and it models negative connections between vertices, very useful in a social setting. In fact it can detect cliques within a social network.

It turns out that the calculation behind pagerank can be achieved by a network of cooperating peers none of whom have a complete view of the network.

If we couple this calculation of abnormalities with feedback at each peer, we could create a swarm of peers that actively shuns any peer which is disproportionately important. Sort of like a swarm of microdrones doing formation control (also lots of distributed algorithms).

Perhaps each peer could calculate the centrality of the IP blocks of the peers it knows, thus the network as a whole would be able to detect that malicious entity that runs too many peers.

Our graph of vertices can be different that the actual connection between peers. For example, in a DHT we could do the centrality calculation on the data and its links to other data, to find the importance of each data piece. Or on the accounts/identities, topics, votes and comments of a discussion platform. In this case the feedback mechanism could act to shun identities or cliques with biased voting habits.

Essentially, what I'm imagining is a network of peers that together provide a distributed platform and also police the structure and functioning of the connection network and the platform network of data.

[–] phoxy 1 points (+1|-0)

Decrentralization is good but distributed computing is the future, in my opinion. Email, the Web are arguably decentralized, for example, while Aether, Bitcoin, Ethereum are distributed. Many decentralized projects like Owncloud, gnusocial, diaspora, are really federated- they still require a server, but anyone can set up a server and the reachable network extends beyond any one server, exactly like email and the web.

Distributed computing is what drives bittorrent's magnet links. No one can take them down because the data is distributed amongst the peers that make up the swarm, as long as the swarm exists, the data will be there. The data is copied and reallocated to different peers as the swarm grows and shrinks.

A big problem with current distributed computing (DHT based projects, like magnet) is that a malicious entity could subvert the system if it can control a significant number (majority) of peers in the network. The core problem is that an identity has power (rating discussions, voting, popularity ranking) within the rules of the system and that identity creation is anonymous and trivial. Sort of like how mass unrestricted account creation thwarts Voat's account restrictions by upvote farms. Most attempts to make identity creation harder end up with centralization on an authorizing entity, like a certificate authority. Some, like the blockchain, require proof of work, largely a waste of computing power.

More and more algorithms are being developed which enable complex behaviour by a swarm of simple nodes. The hashgraph algorithm is exciting in that it claims to solve the byzantine generals problem (ensuring success in an unteliable network with potentially malicious peers). It is similar to the blockchain, the peers create a signed, encrypted message to add, and gossip about the addition with other peers and, crucially, they gossip about the gossip from other peers. Eventually, each peer is able to verify that all peers know about all messages, and they all reach consensus on the order of messages.

Meshnets:

I can't see meshnets replacing ISPs anytime soon. ISPs exist to connect to high bandwidth low latency backbones and meshnets are exactly the opposite: low bandwidth, high latency.

The physics of a meshnet are a not insignificant obstacle to low latency operation, too, because a message must be routed amongst more peers than in a backbone network. There are fancy distributed algorithms for constructing spanning trees that could create optimal low hop paths through the network, but still the range of a wireless peer is less than a wired ISP network; low range = more hops = high latency. Wireless connections also suffer from interference when in high densities, already problem in apartment blocks with wifi.

Bandwidth for meshnet peers is also a problem, but perhaps less so since many paths can be chunked together when transferring information. The main problem is that the bandwidth of a path is capped by the slowest peer in the path. Again, distributed algorithms could come to a consensus on optimal routes involving the high bandwidth peers to achieve better throughput. But then these algorithms basically recreate backbones through the peer network.........

Where meshnets might work is in communications which are primarily local. If 99% of the information you access is within the city, then the number of hops to get to it could be low.

[–] PMYA [OP] 2 points (+2|-0)

Good answer.

Distributed computing is what drives bittorrent's magnet links

This is what I meant, I mixed up my jargon.

low range = more hops = high latency

I didn't think about this. That is a problem I don't have an answer to. Maybe the answer will relate to new ways of transferring data. We are still in the infancy of this technology, and if shit like this is possible, we can't even imagine what might be the standard in 20-30 years.

[–] phoxy 0 points (+0|-0)

Maybe the answer will relate to new ways of transferring data. We are still in the infancy of this technology, and if shit like this is possible, we can't even imagine what might be the standard in 20-30 years.

Absolutely. If we can develop a cheap common high bandwidth (Gbps to Tbps) interconnect then meshnets become much more practical.