Freemors Blog

Musings of an East Coast Techie
Posts tagged as learn

Three Types of Internet

2016-01-07 by Freemor

I am writing this blog entry to explain to those that may not know the three models of doing things on the Internet. And also why it is important to understand them, to pay attention to them, and choose services and software that use the most correct model.

The three models are:

They all have their strength and weaknesses and more importantly, they all have an impact on your rights and freedoms.

Centralized

Centralized is the most common. This is your Google, Facebook, Pinterest, Twitter, Amazon, E-bay, Bank, Etc.

The centralized model has Big servers run by private interests (the site owners/Company) located in some place of their choosing which you use a browser or mobile App. to connect to. Typically all data is stored on the remote server.

This model is perfect for things like banking or online shopping. Just like in the real world you go to the place of business to shop or bank. It is also very appropriate for information type websites news,stocks,weather,sports scores,etc.

The important thing to remember about this model is that you do not control the server and therefor you do not control the data on the server. For the sites mentioned above, no biggie. For things like Facebook,Twitter,etc that live and die on user generated content (your stuff, your data) it's a huge biggie. Once your data is on their server it is usually considered "their data". The User Agreements of such site almost always stipulate that they can do what ever they want with what you upload.

The centralized model is also the easiest for the government to spy on, sensor, control, and shutdown. Because all the data on the server is owned but Company X all the government has to do is legally compel Company X to hand it over. In this way encryption like HTTPS is null and void. Governments can also just seize and shutdown servers they don't like. Also if Company X gets tired of running the server it and all your data will just go Poof and disappear from the Internet.

Considering all these things it is easy to see that the centralized model is both the least free (as in your rights and freedoms) and the most fragile. A lot of service providers out there could be whipped off the Internet by one good flood or other disaster happening to their main server.

Federated

This model is less known and understood by the average person today but it is actually the most common model used in the early days of the Internet. In this model instead on one server (or server farm/s) owned by one company there are many small servers that all talk to each other (federate) to provide a service. This model is used for E-mail, IRC, Usenet, XMPP, UUCP (yes I know that is ancient and deprecated), and newer system like pump.io and Tor. The strength of this system lies in the fact that no one owns the system.. sure they may own a server or two but no one owns the whole system. If a server goes down you just switch to another one.

This model is much harder to sensor, shutdown or control. Servers can live in different countries with different laws and governments. Typically the software to run these kinds of servers is small and easier to install and maintain. This means that anyone with a bit of work and understanding can set up an server and become part of the network of servers. If a government wanted to shutdown the service they'd have to block access to every single server, or a majority of them, to make the system unusable. Not so easy. Spying wise it is harder too. If the government compelled Google to hand over all E-mails (you can be pretty confident that they have/are) it doesn't get them any mails going from sue@freeserver.org to bob@ilivefree.net.

Users typically use some sort of "client" software to connect to their server of choice and interact with the system as a whole. They don't have to worry about what server their friend is on because all servers in the system talk to one another. So sue@freeserver.org can email bill@gmail.com no problem, no worries. As you can see from that example the part that comes after the @ actually refers to what server someone is on in the system. The same is true for XMPP addresses, SIP (proper Voip) addresses, webfinger addresses (pump.io), etc.

There is still the problem of your data on their server.. but as a federated system passes the data from server to server people running federated servers tend to act more like custodians of the data then owners of it. People tend to run these types of servers to offer a public service. OK, well not the Google's of the world. But places like Riseup or Ostel.

Peer to Peer (P2P)

In this model the client software is also the server. All clients on the system talk to and can connect to all other servers on the system. These systems are highly dynamic (servers coming and going all the time) and tend to be very connection and bandwidth heavy because everyone has to help move everyone else's data around.

In a P2P system no one owns the data it just lives out there bouncing from client to client. This means that for most P2P systems you have to be willing to give resources to the network. You have to let the P2P network use some of your bandwidth and disk space.

As you can imagine this is the least easy to censor or shutdown model, and also, if it is done right the hardest to spy on. Because of this many people see the P2P model as a freedom and privacy Panacea. But the truth is this isn't the best model for all things. I don't want to be trading huge chunks of bandwidth and disk space just to see what the weather is going to be like tomorrow. Also because of the dynamic nature of the network and the problem of where stuff is stored relative to who is online the P2P model isn't really the best for "store and forward" applications like E-mail. Sure there are things like Bitmessage but if Bob isn't around for a day or two after Sue tries to send him a bitmessage her software will have to try sending it again. If they have really bad timing it could take months for Bob to get the message. Where in a federated system Sue would send the data to her server of choice which would send it to Bob's server of choice which would hold on to it till bob came online.

People in remote locations or developing countries may not have the bandwidth or disk space to share. There are people in the area where I live for whom a P2P system could easily eat their monthly data allotment in a day or two.

Even tho a P2P system that used good encryption for transfer and storage would be very hard to spy on these systems are complicated beasties and are prone to other forms of attack, resource depletion, evil clients that do things like say they'll forward that data but then throw it away thus vanishing it from the network, governments running a ton of clients to analyze the traffic flow and figure out who is talking to who or even who is who, etc.

It is also important to note that many P2P systems like Bittorrent and Bitcoin do nothing to hide your IP address, so there is no anonymity. Many people are confused and think that P2P automatically means anonymous.

Which is Best

There really is no one best model. The important thing is to try and pick the services that are using the right model for the right job and be aware of the trade offs

Things to watch out for are centralized sites that are trying to own and control your data, and a newer trend of big companies trying to push the workload onto users by using P2P technologies. Netflix has eyed this to take some of the load off their servers by making people watching a show also stream that show to other people watching the show.. great for them.. terrible for your bandwidth.

Pay attention to which model a service is using and you have a much better ideal of how it effects your rights, freedoms, data, bandwidth, and disk space.