Comment by unethical_ban
1 day ago
I don't knock Gemini for existing and being a neat project, but even for hobby it seems too restrictive. No cookies means no authenticated interaction with a site, no inline images means it's less informative than a 100 year old encyclopedia.
Perhaps a "Simple Web" spec could be created to audit a site and verify its privacy and simplicity protections. Things like "Cookies only for auth", "No JS" or "low JS", "No ref tracking in or out", "No tracking pixels", etc.
You'd have to prove these things are possible in the face of the ingenuity of the entire adtech industry. The limitations you point out, on the other hand, have easy solutions:
* auth: Look at https://github.com/kr1sp1n/awesome-gemini#services Tons of services support some form of auth.
Edit: https://martinrue.com/station is another service I use that's missing in the above list.
* images: click to load
Janky but doable. Janky is the price you have to pay to avoid adtech.
What makes you think the adtech industry won't just embed ads in text if this becomes popular at all. You don't need any kind of technology for ads, only eyeballs.
Adtech isn't about showing ads. It's about targeting ads. And yes, if Gemini gets wildly popular maybe they'll think of something. Good problem to have! New mistakes instead of same old ones!
> Janky is the price you have to pay to avoid adtech.
I don't understand, unless adtech is holding your family hostage and forcing you to adtech. Can you elaborate?
If you want to support cat pictures that show up without clicking a link, but prevent any behavioral exhaust from tracking pixels, that seems to be an open problem. Every new feature is like this: a risk surface until proven otherwise. So to reduce risk you have to limit features, i.e. jank.
3 replies →
Authenticated sessions are supported using client certificates.
No inline images is a significant restriction indeed but it also gives you a high degree of confidence that most Gemini pages will be very lightweight. I don't find it that limiting. It all goes back to the point that Gemini is intended to supplement the web and not replace it - if you want image heavy content you can get it elsewhere. Personally I find the lack of inline formatting and links more frustrating.
No inline images is not a restriction of the protocol. It's a restriction of the text/gemini MIME content type, and of the browser implementations. A server can still respond with text/html content over the GEMINI protocol, with embedded <image/> data. The GEMINI protocol specification does not restrict what RFC2045 types can be used.
* https://geminiprotocol.net/docs/protocol-specification.gmi
* https://iana.org/assignments/media-types/media-types.xhtml
What exactly do you think you are proving? It doesn't matter what some spec theoretically allows, if you want to server something to gemini users it can't have inline images.
Nothing prevents a gemini browser from showing inline images (though it might be officially discouraged?). They are just links.
But actually loading images separately can work well. If you are reading for the text content you can save the time and bandwidth to load of all the images, or maybe you want to look at one image in detail, you can load just that one, and zoom or frame that independently of the surrounding text.
I do think it's a shame that Gemini doesn't have images and richer text, but maybe it would be even less popular/successful if it had those. Gemini won't be the last of these simple protocols so it's a useful learning opportunity.
My project at the moment is kind of related to these "simple web" ideas. Instead of giving up on HTML altogether I'm making a simple web browser, to see if there's a way to render even relatively complex existing pages, like Wikipedia or news sites, without needing to implement much or any CSS. A bit like "reader mode". (link if you are interested: https://codeberg.org/kaimac/weaver)
Inline images are a client implementation/styling detail. Some clients have it, but most don’t as most users don’t seem to want it. I believe Lagrange (the most “visual” browser) has this feature.
It's https://smolweb.org/
How about:
- no scripts of any kind
- no cookies
- no forms
- all resources (e.g., styles, images) needed for display inlined
- a spacious minimum cap on data URI length
- elaborate the <a> tag a bit to allow a series of content addresses (hashes, IPFS, magnet URIs, etc.) for references
Basically, a "dead" subset of HTML suitable for distributing documents.
I keep writing the same comment every time this is brought up, but browsers need to support text/markdown.
Technically depending on the implementation, markdown is a superset of HTML.
You'll need to be more specific since there are many variants of Markdown and the original explicitly permits arbitrary html.
2 replies →
Markdown? Terrible "spec".
Browsers already support XML.
You can spin up a HTML-but-restricted XML grammar (with extra stuff even, like footnotes and stuff) and a CSS file in maybe half an hour, and it'll render in your browser just fine.
(Yeah, it'll be missing all the accessibility provisions, but you know, the base to build on is there, whereas "MarkDown in the browser" rendering has been often suggested and never implemented).
Sooner or later "inline images" == "advertising".
And "tracking pixels".
Keeping them separate was a smart move, and entirely consistent with the underlying philosophy.