Authentication for AI Catalog HTTP Endpoints

[LukasHeimann]

The current spec (and also a2a/mcp server cards it builds on) always seem to assume the HTTP endpoints they propose (under .well-known or elsewhere) to be unprotected and readable without authentication.

In enterprise scenarios, that is typically not desirable – you'd not want to expose metadata on your corporate AI assets publicly. The MCP Registry standard proposes, for those cases, the same authentication scheme as MCP itself, i.e. OAuth and discovery via WWW-Authenticate, OAuth Protected Resource Metadata, CIMD, ... – https://github.com/modelcontextprotocol/registry/blob/main/docs/reference/api/registry-authorization.md

Are there already discussions to recommend the same mechanisms for the a2a/mcp server cards, ai-catalog and ard ecosystem as well? By being in the standard, the likelihood of client tools (and SDKs) integrating authentication support would be vastly improved.

I have tried to look but am honestly not too sure which of the various standards (or all?) would have to mention authentication...