There’s a niche for simpler/leaner web technology.
Firefox and other browsers have a “reader mode” that simplify a page to help focus on the content. Google tried something with their own AMP service.
Gemini deserves some credit for writing a spec and for trying, even if it has many issues. I hope Gemini improves based on this kind of review and critics.
My bet is on a subset of modern web tech (eg HTTP3 + TLSv1.3 + HTM5) so client software don’t have to strugle with compatibility and legacy protocols/formats, while allowing access for classic browser users. It could explitiply exclude or discourage some features (JS, WebUSB, GIF, autoplay of video, frames, 3rd party cookies, ping attribute…). This would limit distraction, better protect privacy, and hopefully keep the implementation simple.
Gemini deserves some credit for writing a spec and for trying, even if it has many issues. I hope Gemini improves based on this kind of review and critics.
Does it? Seems like its just Gopher’s spec rewritten by people with very little knowhow of how you write a spec properly. Only thing they seem to have done competently is the spec for the page rendering-format (…which they accomplished by copying Markdown… and which Daniel points out, should not be part of the protocol-spec), everything else in the spec seems like every client YOLOs.
Yes, they do because it’s feeding the discussion on leaner Internet technology.
You’re right in that Gemini has many issues and isn’t viable as-is. It’s clearly incomplete and early-draft quality, lots would need to be changed based on implementation feedback, including Daniel Stenberg’s feedback.
As mentioned my bet is on a subset of web tech, not Gemini. Gemini appear right now to be a playground for implementation to work on UI, markdown use as HTML replacement, … but doesn’t provide a viable protocol.
Not to make it all about personalities, but “this HTTP client person” in this case is the lead dev behind probably the most heavily used HTTP client library of all time. (Not even sure where I would begin to find those kinds of statistics, but I’d be shocked if something else came anywhere close to cURL’s usage.)
I mention this not because his opinion is infallible but because he’s seen some weird edge cases and weird use cases over the last quarter-century or so.
