Rebooting the Internet from Orbit

Abstract

Low Earth Orbit (LEO) megaconstellations are the first network in three decades being built greenfield at a planetary scale, and the window for shaping their architecture is closing. Once operators lock their inter-satellite and ground-gateway data planes to IP conventions, the field will have lost its last realistic opportunity to deploy a non-legacy inter-domain architecture without coordinating a billion endpoints. The terrestrial Internet has resisted architectural change for two decades, as IPv6 illustrates: standardized in 1998, it crossed 50% on Google's measurement only in 2026. Clean-slate designs such as ICN, SCION, and XIA, and extensibility frameworks such as the Trotsky Processor (TP) and Plutarch, have not yet broken this rigidity. We argue that non-terrestrial networks (NTNs), in synergy with 6G, offer the last realistic substrate for the long-anticipated revitalization of the Internet's architecture, and that TP is the most credible primary vehicle for that reboot, complemented at the edges by translation approaches. We sketch two concrete blueprints, with TP either native to the orbital segment ("Trotsky in the Sky") or anchored on the ground and using the NTN as a globe-spanning logical pipe between L3.5 nodes ("Trotsky in the Trenches"). A capacity analysis of the Starlink constellation shows that the V3 generation entering service in 2026 makes capacity a non-issue: fewer than 2K V3-class satellites would absorb today's entire yearly inter-domain traffic, well within the ≈10K currently in orbit.