問題文
A team wants distributed tracing across their MCP calls. Which mechanism does the specification document for this?
選択肢
- There is no documented mechanism; teams must define their own vendor-prefixed keys.
- The reserved _meta keys traceparent, tracestate, and baggage, whose values follow the W3C trace context and baggage formats.
- The logging feature, which is the specification's recommended observability mechanism for carrying trace identifiers between hops.
- A dedicated tracing capability that both sides negotiate, after which the server emits span notifications on the subscriptions/listen stream so that the client can assemble the trace without having to correlate anything itself.