Peter Alexander

So yeah, the correct header is `Mcp-Session-Id`, also the session ID should be generated by the server as part of the initialisation handshake, it's not something that the client should...

> POST a JSON array of its buffered responses Is this a new RPC? We don't have batching any more. > complex state correlation mechanisms across instances (even when shared...

It's interesting that the original [Streamable HTTP transport introduction](https://github.com/modelcontextprotocol/modelcontextprotocol/pull/206) intended for it to be optionally usable in a stateless manner. > A completely stateless server, without support for long-lived connections,...

Thanks for putting this together @tulshi, a few thoughts: 1. If we're going to specify a version format, why not https://semver.org? I don't really have strong opinions on what is...

This makes sense for some forms of extensions, but one of the reasons I somewhat punted on specifying negotiation approaches for extensions is that I'm not sure this really works...

> Why allow servers to add unrequested extensions? A problem with this is that it makes extension negotiation a multiple-round-trip process. In a stateless setup, the extensions the server sends...

Good catch, looks like a typo, but this change should be made to the draft spec. I don't believe we retroactively update already published specs.

> Out of curiosity, are new spec versions not recommended to have breaking changes? I understand the line of thinking of not changing specs that have been finalized and released...

> MCP servers supporting HTTP-based transports (including Streamable HTTP and SSE) SHOULD provide a server card via a .well-known URI. Why not "MUST"? Feels like this could be a very...

> I think i want to give developers the choice if the opt into discoverability or not. I am probably okay with saying that you MUST offer it as a...