At IETF 126 in Vienna, the agentproto BoF held its chartering vote this week, working from groundwork laid by a preceding gateway-focused BoF, DMSC. The stakes are unusually concrete for a standards process: rather than starting from scratch, DMSC spent its session mapping where the AI agent gateway model already has coverage and where it doesn't. That gateway model is worth naming precisely. It's the architectural layer sitting between an orchestrating agent and whatever downstream agents or tools it invokes — the point where one agent's request has to get routed, authorized, and reconciled against parallel threads before it lands anywhere useful. A gateway has to handle capability exposure, request forwarding, coordination and synchronization, policy control, observability, and secure inter-agent identity verification, all at once. That's a lot to ask of a layer that today is assembled ad hoc from whatever OAuth, MCP, and A2A happen to provide. The DMSC group's actual contribution was diagnostic rather than prescriptive. It profiled how existing mechanisms — defined separately in OAuth, WIMSE, DNSOP, and the A2A and MCP specs themselves — already cover pieces of that six-part problem, and flagged where nothing covers it at all. That gap analysis was built specifically to sharpen the chartering discussion at agentproto. The framing document behind agentproto (the Rosenberg-Jennings Internet-Draft) treats AI agent protocols as their own new application layer, distinct from the web and email layers the IETF has organized around for decades. An earlier IETF blog post from January laid out five unsolved questions no current protocol fully answers: how agents discover each other's existence and capabilities, how they delegate tasks and credentials across organizational boundaries, how user privacy survives context handoffs between systems, how multimodal content gets handled as context accumulates, and how humans stay in the loop before irreversible actions commit. What makes this notable isn't that a BoF happened — plenty of BoFs don't go anywhere. It's that the working group being voted on this week is explicitly scoped around coordination and routing failures that MCP and A2A, as deployed today, don't resolve on their own. If it charters, it becomes the first IETF effort aimed squarely at standardizing how one agent hands work to another across organizational lines, rather than leaving that to whichever vendor's gateway product wins the market first. The vote's outcome — whether MCP and A2A harden into de facto defaults or yield ground to an open standard — will shape how agent-to-agent handoff gets built for the next several years.