I understand the protocol is simple, but there's no reason the client has to be. Taking your example:
> CSV files become inline tables if you expand them, etc.
Why not only make them inline tables, but make them searchable, pivotable and graphable? Embed a bunch of statistical tooling directly in the interface. It's something I dream would happen on the Web - if Google^H^H^H^H^H^Hweb standard bodies could agree on a new tag for tabular data, say <datatable>, that would - by default - come with aforementioned capabilities, and with enough styling and hooks to lure in the webdevs, then maybe all the tabular data that's on webpages could be finally made explorable in an efficient fashion...
I hear you, and I definitely imagined something along those lines eventually wrt. basic analysis tools in-browser.
In my mind, the key idea is that instead of inventing a new standard or extending an existing one (a new HTML tag, for example), what we need to do is compose formats which are each individually simple.
CSV is far from perfect, but it has an RFC, it's simple, and there are already loads of tools that can work with it. Many times I have wanted to work with data that was displayed in a tabular fashion on a web page, and I have variously used copy-and-paste and manual cleanup or writing a scraper to get it into a workable form. Imagine how much easier my life could have been if those tables simply were CSV files that my browser happened to display nicely, and I could have fetched them directly at any time!
A format like text/gemini is an opportunity to start from scratch with this kind of compositional thinking in mind. Smart browsers could do all sorts of convenient things, but even "dumb" browsers and scripts benefit from having data and content in accessible formats.
> CSV files become inline tables if you expand them, etc.
Why not only make them inline tables, but make them searchable, pivotable and graphable? Embed a bunch of statistical tooling directly in the interface. It's something I dream would happen on the Web - if Google^H^H^H^H^H^Hweb standard bodies could agree on a new tag for tabular data, say <datatable>, that would - by default - come with aforementioned capabilities, and with enough styling and hooks to lure in the webdevs, then maybe all the tabular data that's on webpages could be finally made explorable in an efficient fashion...