Brief post, but I was reading up on the Spartan protocol in the spec and noticed something that raised my eyebrow.
The spec uses text/gemini for page response but extends gemtext to include this:
=:[<whitespace>]<URL>[<whitespace><USER-FRIENDLY LINK NAME>]
This is a way to do prompts that differs from Gemini's 10 INPUT response. I have no opinion on this other than it should *not* use text/gemini as the mime type.
A mime type should not mean different things depending on which protocol it is coming from. Imagine if png behaved liked gifs when served over gemini://. It would just be bizarre.
Imagine you open a .gmi file in your text editor. Does the editor support the =: or not? It has no way of knowing which protocol it will be served under. Imagine opening said .gmi file in a browser with the file:// protocol. Does it display the =: as Spartan would want?
The only reasonable thing to do here is fork text/gemini and define it as a new mime type text/spartan and provide a new file extension recommendation for people to use.