the second realm

i2p is often called the invisible internet. the name points at something real, then asks too much of it. a network can limit what its participants learn from a route. it does not erase every fact about the path.

a second map

i2p puts a second map beside the familiar one. applications address cryptographic destinations. routers carry messages between them. the destination is an application endpoint, not an ordinary public host name with a different suffix.

that distinction changes the first question. instead of asking where a server lives on the public net, the application asks how to reach a destination through the network currently available to it. sites, mail, chat, and files can live there. the map is part of the service.

the route has a cost

i2p messages travel through inbound and outbound tunnels. the separation is deliberate. latency and throughput are part of the trade.

those costs are not defects around an otherwise ordinary web service. they describe the service. a network built around indirect routes gives up some immediacy and asks more of its participants. the bargain belongs in the design, not in a footnote.

a bounded promise

transport has a narrow job. it can limit which relationships are exposed by the path. identity, custody, content, and trust remain separate jobs.

a destination does not establish who operates it. a hidden route does not protect a private key. encryption does not decide whether the person at the far end deserves trust. each boundary still needs its own answer.

the second realm has its own addressing, routes, and failure modes. this log will return to that map, and to the boundary between a privacy claim and proof.