Addressing Communication Problems on the Application Layer

The issue I have with the term "mesh networks" is that it is associated with a flat network topology. While I love the idea of avoiding hierarchy, this simply doesn't scale.

Data Plane: How the Internet Scales

The internet on the scales because it has a tree like structure. As you can see in the diagram below, global (tier 1) ISPs branch out to national (tier 2) ISPs who in turn branch out to local (tier 3) ISPs.

                    ,-[ Tier 1 ISP (Global) ]─-───────[ Tier 1 ISP (Global) ]
                   /           |                         
                  /            ▼                         
          [IXP (Global)]═══╦═══[IXP (Global)]             
                           ║                              
                 Tier 2 ISP (National)◄──────────╗               
                    /       \                    ║               
                   ▼         ▼                   ║               
         [IXP (Regional)]════╬══[IXP (Regional)] ║               
                 /            \                  ║               
                ▼              ▼                 ║               
     Tier 3 ISP (Local)  Tier 3 ISP (Local)      ║               
              |                |                 ║               
              ▼                ▼                 ▼               
           [User]            [User]         [Enterprise]

▲ IXPs are physical switch fabrics - members peer directly
▲ Tier 1/2 ISPs provide transit through IXPs but don't control them

This structure also reflects in IP addresses, where regional traffic gets routed by regional tiers and global traffic keeps getting passed up through gateways till it reaches the root of the tree. The global ISP then routes the traffic into the correct branch so that it can trickle down to the destination IP at the bottom.

12.0.0.0/8           - Tier 1 manages routing (IANA-allocated)
└─12.34.0.0/16       - Tier 2 allocated block (through RIR)
   └─12.34.56.0/24   - Tier 3 subnet via upstream provider
      ├─12.34.56.1      Public IP (CGNAT pool)
      └─192.168.1.1     Private IP (local NAT reuse)

Balancing idealism with pragmatism

This approach to scaling is much less idealistic than a flat hierarchy, because it relies on an authority (IANA) to assign the IP ranges to ISPs through Regional Internet Registries (RIRs). Even if this authority wasn't required, the fact that many users rely on few Tier 1 ISPs means that the system is inherently susceptible to sabotage (see 2019 BGP leak incident).

Control Plane: the internet is still described as decentralised

The internet is still described as decentralised because there is a flat hierarchy between tier 1 ISPs at the root of the tree.

INTERNET CORE (Tier 1 ISPs)
----------------------------

  AT&T (AS7018) <══════════> Deutsche Telekom (AS3320)
    ║   ╔════════════════════╗   ║
    ║   ║                    ║   ║
    ╚══>║ NTT (AS2914)       ║<══╝
        ║    ║    ║          ║
        ║    ║    ╚═══════> Lumen (AS3356)
        ║    ║
        ╚════╩════════════> Telia (AS1299)

The border gateway protocol (BGP) is used to exchange routing information between autonomous systems (ISPs). Each autonomous system is a branch of the "internet tree" and each autonomous system advertises routes to downstream autonomous systems (branches). However, the autonomous systems at the root of the tree also maintain a record of their piers, so that they can forward traffic to the correct peer. Hence, the following is a more complete diagram of the internet:

              INTERNET CORE (TIER 1 MESH)  
              ╔══════════════╦═════════════╦════════╦═════════════╗  
              ║              ║             ║        ║             ║  
            AT&T (AS7018) <══╬══> Lumen (AS3356) <══╬══> NTT (AS2914)  
              ║              ║                      ║          
              ║              ║                      ║          
              ║              ╚══> Telia (AS1299) <══╝        
              ║                               ║          
              ╚═══════════════════════════════╝          
              ║                               ║          
              ╚═> Deutsche Telekom (AS3320) <═╝        
                    ║                        
       TREE HIERARCHY BELOW - MESH ABOVE  
                    ║                     
					▼  
             [ Tier 1 ISP ]───────────────────┐  
          /           |                       |  
         /            ▼                       ▼ 
       [IXP]═══╦═══[IXP]                    [IXP]  
               ║                              ║  
               ▼                              ▼  
         Tier 2 ISP◄──────────╗           Tier 2 ISP◄───────╗  
        /       \             ║            /     \          ║  
       ▼         ▼            ║           ▼       ▼         ║  
       [IXP]═╦[IXP]           ║          [IXP]═╦[IXP]       ║  
        /     \               ║           /     \           ║  
       ▼       ▼              ▼          ▼       ▼          ▼  
       Tier 3 ISP            User       Tier 3 ISP      Enterprise  

So its a mesh network - whats wrong with that?

In the example above, NTT can only send traffic to Deutsche Telekom via Lumen or AT&T. NTT relies on its peers to maintain a correct record of the IP range that they are responsible for, so that traffic that was intended for Deutsche Telekom doesn't end up in the wrong network. An intentional or even accidental error in an autonomous system's routing tables can be detrimental to the flow of traffic through the network.

Hence, mesh networks require:

• Reliability: peers rely on each-other not to fail (e.g., 2019 AWS US-East-1 availability)
• Trust: peers must be honest about address ranges (e.g., 2018 BGP hijacking incident)
• Central planning: BGP traffic engineering determines which route a packet takes

Application Layer Innovations

Now that we have a rough overview of how the internet is broken, lets think about what can still be done. The flat hierarchy that we associate with mesh networks sounds beautiful, but it doesn't scale. However, the tree-like structure assumes that most participants in the network rely on an authority to give them an address or a range of addresses so that they can communicate.

Overlay networks

Fortunately all it takes to interact with someone on nostr is their public key. The recipients client will render your signed and/or encrypted event no matter how it reaches them. Whether your note reaches them over the internet, over some other network or via a carrier pigeon doesn't matter.

is working on a NIP for that and I'm sure he will share an explanation of how it works.

Bitcoin Instead of a Routing Algorithm

Now that we have digital bearer assets (e-cash), users can pay their internet gateway (TollGate) for access to the internet even though they are still offline. Once the TollGate has redeemed the e-cash, it gives the user access to the internet.

Frictionless Switching between ISPs (TollGates)

Legacy internet service providers use KYC money (fiat) to transfer the cost of the infrastructure to their users. However, this means that they are able to identify which packet belongs to which user. The KYC nature of their interaction with the users also makes it difficult for users to switch ISPs when service providers undercut each-other. Internet users who are on e-cash rails can hop between ISPs frictionlessly since they buy small amounts of data frequently in a granular manner.

Users that have access to independent competing TollGates can switch between them freely, so its impossible for any one TollGate to prevent a user from connecting to the internet. The only thing a TollGate can do is attract traffic by providing cheaper and/or faster internet access.

Anyone can arbitrage connectivity

Now that users have non KYC internet, there is nothing stopping them from reselling access to their internet connection. Anyone who has a WiFi router and access to a cheap internet connection can act as a range extender by re-selling access to that gateway for people who aren't able to connect directly.

Now the route that the traffic takes through the network is determined dynamically by the markets. The individual TollGate operators select their gateways and set their prices when they create a business model. The customers select the route that best meets their needs by selecting a gateway for their next purchase.