A new preprint proposes something DNS-based discovery schemes have struggled with: letting an agent's identity survive when it moves. The paper's framing is blunt — multi-agent systems face a fundamental architectural flaw: agent identity is bound to network location. When agents migrate between providers, scale across instances, or federate across organizations, URL-based identity references break, audit trails fragment, and someone has to coordinate the fix by hand. The proposed fix is a new URI scheme, agent://, built from three independent parts: a trust root that establishes which organization vouches for the agent, a hierarchical capability path that describes what the agent does, and a sortable unique identifier that stays stable even as the agent's network address changes. Discovery happens by querying a distributed hash table keyed on capability rather than location, so a lookup returns agents by what they can do rather than where they currently live. Trust-root scoping keeps a company's internal agent namespace from colliding with someone else's, while still allowing deliberate federation across organizational boundaries. Cryptographic attestation via PASETO tokens binds capability claims to agent identity, enabling verification without real-time contact with the issuing authority — meaning a downstream service can check an agent's claimed permissions without phoning home to whoever issued them. The author backs the claims with numbers rather than just architecture diagrams. The scheme was evaluated across four dimensions: capability expressiveness (100% coverage on 369 production tools with zero collision), discovery precision (F1=1.0 across 10,000 agents), identity stability (formal proofs of migration invariance), and performance (all operations under 5 microseconds). That last point matters for anyone worried discovery infrastructure adds latency to every agent-to-agent call — sub-5-microsecond operations are effectively free next to network round-trip time. This sits closer to a formal proposal than an adopted standard — it's a single-author preprint, not yet a deployed protocol with vendor buy-in. The DHT-based discovery layer also introduces its own open questions, including incentive design for nodes that host the hash table, which the author flags directly as unresolved. But the core insight — that identity needs to be portable across topology changes, not just verifiable at a point in time — addresses a gap that location-bound naming schemes have mostly ignored. Whether agent:// gains traction will depend on whether registries built on DNS-style discovery see enough migration and federation pain to want a topology-independent alternative.