Introduction

The concept is that the address space can be extended from what has been specified. A key insight is that addresses low in the tree structure will encode many null coordinates, and that represents a lot of address space that is not being utilized. This space can be somewhat utilized by specifying that additional, non-null identifiers can follow the first fully null nibble that is not part of a coordinate.

Utilizing more of the address space must be done carefully to avoid breaking the greedy routing algorithms. Because these new addresses will be incompatible with the greedy routing system, nodes with these addresses cannot be parent nodes in the network. I suspect that in practice, only dedicated routers will run the full packet switching node logic, while devices that have other responsibilities/roles will utilize this secondary address type, much the way WiFi devices "lease" addresses from WiFi routers. This should not be a problem as the router hardware will be very inexpensive, and the standard hardware implementation will be open source.

Proposal

Proposed solution:

1. Determine the number of bits required to encode the router's address.
2. Add 4.
3. Subtract that number from 32. If it is zero, the operation fails.
4. Generate that many random bits. Regenerate if the bits are all null or if it collides with an existing, active lease.
5. Concatenate the router's encoded coordinates, a null nibble, and the random bits to generate the leased address.
6. Store the leased address and the relevant connectivity information needed for delivering datagrams.

Routers lower in the tree structure will have more possible lease addresses, so this is theoretically not "fair" in a sense, but the "fair" way to do it is to intentionally limit the capacity of such nodes. The only potential justification would be an argument around incentive to grind out key pairs to get one's node elected as the root for the benefit of having more lease addresses available, but this seems like a very unlikely problem -- unless the network grows to such a large size that the full Pycelium specification must be implemented as envisioned years ago with bridges between autonomous domains, nodes will practically speaking have as many leasable addresses as they could possibly need.