On Thursday, July 23, the agentproto Birds-of-a-Feather session at IETF 126 in Vienna brought a question that's been simmering since January into the Internet Engineering Task Force: should agent-to-agent communication get a formal RFC, the way HTTP or TLS did, instead of staying a patchwork of vendor protocols like Anthropic's MCP and Google's A2A? This wasn't a vote and it wasn't a product launch. As one report on the session put it, this is not a vote and not a product launch; the IETF works by rough consensus, and any resulting standard is years away. What matters is the shift in framing: agent-to-agent communication is now being treated as an internet-infrastructure problem, not just a vendor feature. The session built on real momentum. An IETF 124 side meeting on the same topic drew roughly 125 engineers in person plus a comparable remote crowd — enough that the agentproto BoF arrived at IETF 126 with a Working Group charter already drafted for community review. The organizers, Jonathan Rosenberg and Cullen Jennings, had laid out five unsolved questions back in January that no current protocol fully addresses: how agents discover each other's existence and capabilities, how agents delegate tasks and credentials across organizational boundaries, how privacy is preserved when passing user context between systems, how multimodal content is handled as context accumulates, and how humans stay in the loop before irreversible actions are committed. Security is the pressure test. Real-world prompt injection incidents so far have come through public Slack channels, GitHub issue trackers, and web pages — attack surfaces documented in existing single-agent deployments. None of these incidents involved inter-domain agent cascades, but they point toward the risk profile that multi-agent deployment at scale will create, which is exactly the failure mode a cross-organizational standard would need to close off before it ships. If the BoF's rough consensus holds and a Working Group is chartered, a published RFC is still likely two to four years away. But the fact that a formal charter draft already exists, and that three of the five BoF slots at IETF 126 touched AI agents directly, signals that standards bodies now see uncoordinated agent-to-agent protocols as an interoperability risk worth fixing at the internet layer, not something individual vendors can patch over with bilateral integrations.