{"slug":"registry-listing-vs-runtime-use","title":"Registry listed, protocol compatible, actually used: three different milestones","description":"Apply MachineRealms’ interoperability lessons without treating directory presence as demand or verification.","sources":[{"name":"MachineRealms: published interoperability evidence","url":"https://machinerealms.com/api/v1/research/interoperability"},{"name":"Model Context Protocol specification","url":"https://modelcontextprotocol.io/specification/latest"},{"name":"A2A protocol specification","url":"https://a2a-protocol.org/latest/specification/"}],"sections":[["A listing is a declaration channel","Publishing an entry can make a service easier to discover. It does not establish that the endpoint is reachable from every client, that the advertised operations work, or that an external participant found it useful. Keep registry publication separate from runtime evidence in both product reports and marketing claims."],["Compatibility is specific","A client may implement only part of a protocol version or expect a particular discovery sequence. A compatibility test should name the client, version when known, transport, operation, timestamp, and result. A successful exchange with one client is not universal compatibility. A rejection can be a useful observation without proving that either party violates the entire standard."],["Record the full sequence","The research record should link declared capabilities, attempted discovery, the operation requested, the observed response, and any repair. Keep an operator’s explanation labeled as an inference unless corroborated. This produces implementation evidence other developers can use, rather than a vague claim that integration works."],["Do not count your own tests as demand","Deployment checks, steward tests, and invited demonstrations should carry their own traffic class. Organic discovery, voluntary participation, and completed contributions are later stages. A funnel can legitimately show successful technical access and zero downstream participation; that is a product finding, not a reason to relabel test traffic."],["Use both products for different jobs","MachineRealms provides the participation and structured research context. AIWebSignals provides observation and interpretation of supported evidence. This observatory links those roles without asserting that one service’s declaration verifies the other, and without copying the same article across both domains as separate original research."]],"next":"benchmark-methods-and-unknowns","published":"2026-09-26","updated":"2026-09-26","status":"published","author":"AIWebSignals Research","url":"https://aiwebsignals.com/research/regulated-commerce/registry-listing-vs-runtime-use","example":{"type":"synthetic","heading":"Three milestones, not one success counter","text":"A fictional registry accepts a service declaration. One identified test client later retrieves its capability document, but no external contribution is observed. These establish a listing and a specific compatibility test, not organic demand or completed participation.","exercise":"Keep declared availability, tested compatibility, organic arrival, accepted contribution and useful outcome as separate stages. Use a receipt or acknowledgment to establish a completed handoff rather than counting the outgoing link alone."},"methodology":"https://aiwebsignals.com/research/regulated-commerce/methodology"}