Conventionalise using accept header to choose document

I saw you closed https://github.com/AnswerDotAI/llms-txt/issues/40 without allowing further discussion. Could we open it for discussion?

it may be a different document, but isn't that what the accept header is made for? it allows letting the server know what the type of content the client wants.

I am leveraging the accept header a lot these days and chose to provide markdown by default if no accept header and no extension to the filename is given. I think this really improves upon the ux

example: https://x.com/janwilmake/status/1912473840439161166

ive explained my thinking here: https://github.com/janwilmake/user-agent-router / https://x.com/janwilmake/status/1904951203340967987

I feel like it's good practice if people don't provide an extension to choose the extension for them. This is especially the case for index files where no filename/extension was provided, but also for clean pathnames without extension.

Start by reading issue #40 and the linked user-agent-router discussion to understand the unresolved proposal. No repository files, tests, or entry points are named, so first locate the document-serving behavior and determine the intended rules for Accept headers, missing extensions, and index paths. Done would require an agreed scope and tests covering the selected document behavior.

Written by the indexing model from the issue text.