Free network architecture
The Free Network, in the parlance of net operations, is an Autonomous System – the name couldn’t be more apt. There is a lot to understand here, so we’ll start with some terminology, then we will give you a rundown of the basic network components, and finally we will give a narrative account of the Free Network is going to eat the unfree networks from the inside out. Understand that this is a draft design – consider this a request for comments.
Here is a direct link to the mailing list discussion that this architecture came from.
Logical peer-to-peer is, essentially, what most people nowadays simply call peer-to-peer. That is, when two machines communicate with one another, rather than with some intermediary machine. Yet, just because there is no intermediary machine, does not mean that there is no intermediary whatsoever. The means of communication acts as a buffer. In the material reality of this situation, the transaction is not peer-to-peer, as bits are entrusted to a third party.
The alternative, then, is what we at the FNF call material peer-to-peer, and regard as the key to a freer, more resilient network. This is when data is transmitted entirely via peer-sharing. Materially peer-to-peer communication occurs in networks with a mesh architecture. The Internet began as such a system, but has strayed from its roots. We herald a return to such a system.
FreedomTunnel is software that provides for the automatic utilization of a Virtual Private Network based on IPsec. It allows for secure, logical peer-to-peer communications between any two FreedomBoxes with an Internet connection.
The Mesh Interface for Network Devices
The MIND is what we call our system for distributing connectivity via a cascading series of mesh networks. Layer 0 is a non-mesh home network, Layer 1 is a neighborhood mesh, and Layer 2 is a regional mesh. The MIND allows for secure, material peer-to-peer communications between any FreedomBoxes in the same Layer 1, Layer 2, or Layer 3 mesh.
Layer 0 is a collection of standard, 2400MHz WiFi links between FreedomBoxes and client devices. This is the access layer of the network.
Layer 1 is a collection of 5GHz, neighborhood-scale mesh networks. Each Layer 1 mesh contains one FreedomTower, and as many as 2500 FreedomBoxes.
Layer 2 is a collection of 3650MHz, regional-scale mesh networks. Each Layer 2 mesh contains one or more FreedomLinks, and as many as 250 FreedomTowers.
Layer 3 is a fiber-line implementation of mesh routes between FreedomLinks, and represents the endpoint of the FNF roadmap.
Layer 0 – Client Devices
The Free Network is designed to work with any computing device. Client devices, such as laptops and desktops, connect via ordinary WiFi to a FreedomBox. A thin client application allows them access to all the resources of the FreedomBox to which they are connected.
Layer 0/1 – FreedomBox
The FreedomBox is, in many ways, the most essential component of the network. It is a small, always-on computing device which acts as a router, server, and data storage device. These devices are inexpensive, in terms of both capital and operational costs. Families and individuals can participate in the construction of the Free Network by owning and operating a FreedomBox. In addition to providing Layer 0 connectivity for client devices, FreedomBoxes participate in Layer 1 networking, constantly communicating with other FreedomBoxes that are in range, and ultimately with a neighborhood FreedomTower.
Layer 1/2 – FreedomTower
FreedomTowers, like FreedomBoxes, are owned and operated by those that benefit from them. Unlike FreedomBoxes, however, FreedomTowers benefit entire neighborhoods – communities on the order of five to ten thousand individuals. FTowers perform critical Layer 1 network operations, and help the neighborhood mesh run efficiently. In addition to improving the throughput of Layer 1 connectivity, FreedomTowers participate in the Layer 2 regional mesh. This means that neighborhoods can connect directly to one another, and that material peer-to-peer is not limited in scope to local communities.
Layer 2 – FreedomLink
FreedomLinks, first and foremost, are BGP speakers – they peer directly with other Autonomous Systems of the Internet. FreedomLinks will be situated initially in colocation facilities, and eventually in network meet-me rooms. In addition to serving as primary distribution points for Free Network material connectivity across an entire region, they serve as gateways for logical connections via FreedomTunnel. FreedomLinks will have substantial computational capacity, be multi-homed, and peered via fiber-optics. FreedomLinks are the focal point of the Layer 2, regional meshes – as such, they are intended to serve communities of 25,000-75,000. We estimate that there will eventually be somewhere on the order of 1,000 FreedomLinks in the continental United States, though we are going to start with three.
One of the things that can make the architecture of the Free Network difficult to understand is its emergent, evolutionary nature. It is not possible to simply roll out the type of network described here. Firstly, it is a participatory network, and so it has to be deployed by those that use it. Secondly, the scale and scope are too big to produce and provision the components in one go. So, in order to understand the Free Network, it is important to understand the ways in which it will evolve over time. Presented below are a few of the most probable ways for the Free Network to come into a community.
Of critical importance here is the idea that devices can start participating in the Free Network even before they are able to connect to the MIND and engage in material peer-to-peer. That is to say that if some family buys a FreedomBox to use as their home router, they can begin participating in the Free Network immediately, even if their neighbors haven’t caught on yet, and there’s no mesh to join. This is the purpose of FreedomTunnel. The family still purchases their network access from an ISP, but when they wish to communicate with others on the Free Network, their communications are automatically routed via FreedomTunnel. Eventually, others in the neighborhood replace their routers with FreedomBoxes, and are peered automatically in a Layer 1 network, allowing them to share directly with one another without having to use the Internet at all. At this point, the neighborhood might decide to start an Internet co-op.
Now, in order to understand what an Internet co-op might look like, let’s take a look at another use case. This time, instead of starting with a bunch of Layer 0 networks and moving to Layer 1, we’ll start with Layer 1, and move down to Layer 0. Take, for example, a typical, American subdivision. In this subdivision, you’ve got 100 families, each paying $60/month for a broadband Internet connection. If these families decided to form an Internet co-op they could purchase the same total amount of bandwidth for half of the aggregate price. In order to do so, they would pool their money, build a FreedomTower, and each invest in a FreedomBox. Assuming a price of $5,000 for the Tower, and $100/box, the families would have recouped their investment in five months, and would save $30/month thereafter. As an added benefit, all of their communications with one another would be materially peer-to-peer. Communications destined for outside FreedomBoxes would leave the tower via FreedomTunnel, and due to the asynchronous nature of network usage, individual users would actually see much better throughput almost all of the time.
As the total number of FreedomBoxes in operation goes up, the more routes exist within the MIND, and the less users will have to depend on paid access and FreedomTunnel to communicate with one another. This is the plan: eat the telcos from the inside out – make it possible to start small and grow big, or to start big and use the economies of scale to drive adoption.
You’ve now seen the two basic directions in which the network can grow: up and down. You’ve only got to extrapolate the pattern to see how FreedomTowers motivate the construction of FreedomLinks - in the same way that FreedomBoxes motivate the construction of FreedomTowers. Just as it is possible to build out from a tower, it is possible to start with a Link and build out from there.
Finally, when the Free Network has hundreds of points of presence on the IP core, we can begin to roll out our own fiber lines. This is how we escape the incumbent network operators once and for all. This is the basic premise: we, the members of humanity, are going to build ourselves a network. We will have to pay for the hardware, and we’ll have to pay for upkeep, but in the end we will save untold billions in the process of ensuring our own ability to communicate freely. By becoming our own service providers, we deny the ISPs the ability to profit off of us, choosing instead to provide for ourself, at-cost. Moreover, we will build a network immune to censorship and resistant to breakdown.
Initially, the Free Network will consist of one FreedomLink, and a couple of FreedomBoxes. This is enough for a proof of concept, and is well on its way to realization. From there, it will grow organically. As people adopt the FreedomBox, and get used to the idea of owning their own data, the MIND will grow to meet them, and the Free Network will take form.