Gemini doesn't allow for embedding images in web pages - which makes it vastly inferior to the web for any kind of interesting documents.
Imagine reading a research paper where, in order to view figures and equations, you had to follow a link to a separate object. No font control. No two-column layout. No anchors to allow you to jump to sections of the document. No metadata to inform you of the authors.
Gemini and Gemtext actively inhibit learning and knowledge dissemination by obsessing over pure plain text, which is bad at those things.
This statement is incorrect. Gemini clients can absolutely display inline images.
The difference is that default behavior is to require a user action to load a resource. An image can be a link, but when a user clicks that link it can turn into an inline image. This is how clients like Lagrange work. In other words, inline images can have delayed loading.
If a user understands the consequences (tracking, network usage) of doing so, this behavior can be changed to load images by default; however, authors should not expect users to do this and should write their documents accordingly.
Personally, I prefer a document to not have inline images; my gemini client opens images in my default image viewer instead. My window manager makes my age viewer float above other windows by default. This way, images "pop out" into a separate window that I can keep viewing as I scroll down in a document; I never have to scroll up to look at the last image.
> Imagine reading a research paper where, in order to view figures and equations, you had to follow a link to a separate object. No font control. No two-column layout. No anchors to allow you to jump to sections of the document.
These are all client-side features. Half the point of Gemini is for the user agent to determine presentation and leave semantic markup to authors. I don't want weird fonts or multi-column views, but you do; Gemini lets us both get what we want instead of having everyone see a one-size-fits-all presentation. Clients like Kristall even give you a TOC in the sidebar.
> Gemini and Gemtext actively inhibit learning and knowledge dissemination by obsessing over pure plain text, which is bad at those things.
Text is the only form of communication that can be understood by the sighted, blind, deaf, and machine (translation, etc) while being stored and transmitted without information loss. Text is good at knowledge dissemination.
"clients" and "can" - it's not mandated by the spec, therefore, "Gemini" does not do it.
> The difference is that default behavior is to require a user action to load a resource.
Extremely non-conductive to thought. Again, take the example of a research paper - the difference between having every figure and formula appear by default and having to click-to-load is massive, with the latter being un-ergonomic and inhibiting comprehension and flow.
> These are all client-side features. Half the point of Gemini is for the user agent to determine presentation and leave semantic markup to authors. I don't want weird fonts or multi-column views, but you do; Gemini lets us both get what we want instead of having everyone see a one-size-fits-all presentation. Clients like Kristall even give you a TOC in the sidebar.
You can do exactly this same thing with the modern web with CSS styling and userscripts - the difference being that the web gives you saner defaults that are more conducive to thought, and Gemini clients seem to give you less-sane defaults that are less conducive to thought.
> blind, deaf
This is a limitation of being blind or deaf - someone who's blind wouldn't be able to view a sunset in real life. Obviously, though, while text can be read/listened to by someone who's blind or deaf, that doesn't make text a replacement for images, formulas, or interactive animations - those with those disabilities simply can't perceive the native forms of those things. Several hundred or thousand words describing a layout for a PCB is not equivalent with an image of the layout.
...and, modern webtech has accessibility properties that allow for annotation of non-text media with text. Gemini? Does not.
> machine (translation, etc)
False. Machines cannot understand plain text - it must be parsed. English (and other natural languages) are not machine-parseable, and machine-readable plain text had no reason to not exist as structured data in the first place.
> while being stored and transmitted without information loss.
All of the other kinds of electronic data that exist in the modern web can also be stored and transmitted without information loss, so this is not a special property.
> Text is good at knowledge dissemination.
Relative to text+formulas+images+interactive visualizations? Absolutely false.
Show me how to write out all of the variants of the Schrödinger Equation[1] in plain text, while still making it as readable, understandable, and useful as the mathematical formulas.
Show me how to phrase, in words, a 3D circuit layout, such that it's easier to understand and manipulate than an interactive model.
Show me how to describe the sound of a violin.
Webtech gives you text and images and sound and formulas and interactivity. Gemini gives you text, and that's it. Having to click a separate link to go to a separate object makes it not "part of Gemini" and the user experience is clearly, massively worse.
Prohibiting clients from loading inline images is not mandated by the spec either, so Gemini doesn't prevent it. Loading images inline is fine, as long as it's triggered by a user action. A core idea of Gemini is user control: network requests shouldn't happen without user consent just as presentation should be determined by the user agent.
Non-spec-compliant behavior is also fine if it's explicitly enabled by a user; the default should be spec-compliant.
> You can do exactly this same thing with the modern web with CSS styling and userscripts - the difference being that the web gives you saner defaults that are more conducive to thought, and Gemini clients seem to give you less-sane defaults that are less conducive to thought.
Try changing your browser's default background color and you'll end up seeing a bunch of pages with black text on a gray background. Change your browser's default text layout to two columns and see how many sites still work. The featureset of the web encourages authors to use those features, which begets complexity; complexity begets fragility.
Also, I'm not sure what you mean by "sane defaults"; The "default" HTML presentation is raw markup, and isn't exactly readable. The "default" Gemtext presentation is perfectly readable; in fact, all but two of the blog posts on seirdy.one were initially drafted in raw gemtext rather than markdown. Perhaps you were referring to the default stylesheets of the major browser engines. This is client behavior, and should be compared with existing Gemini clients that focus on presentation as well.
The web allows authors to dictate presentation and deliver content with visual branding; Gemini prevents this to make the focus on content rather than form.
> ...and, modern webtech has accessibility properties that allow for annotation of non-text media with text. Gemini? Does not.
Like the Web and Gopher, Gemini links have display-text. Image links are the same. I consume Gemtext with a screenreader quite regularly, and image consumption is much less painful than it is on the Web. Knowing that users will see text before an image encourages Gemini authors to use good alt-text and to only include images when they convey necessary information that text cannot. Superfluous images are virtually non-existent.
> Machines cannot understand plain text - it must be parsed. English (and other natural languages) are not machine-parseable, and machine-readable plain text had no reason to not exist as structured data in the first place.
That wasn't my point; my point was that text can be parsed and processed by machines much better than other forms of information, improving information dissemination.
> All of the other kinds of electronic data that exist in the modern web can also be stored and transmitted without information loss, so this is not a special property.
Unless you want to load a bunch of 5mb images, you're going to need re-sizing and lossy compression. https://xkcd.com/1683/
> Show me how to write out all of the variants of the Schrödinger Equation[1] in plain text, while still making it as readable, understandable, and useful as the mathematical formulas.
I admit that Gemini isn't great at mathematical formulae. Some people are working on Gemini clients that can understand LaTeX code fences.
> Show me how to describe the sound of a violin.
Include a link to an audio file so it plays when the user wants it to. Several clients can play inline audio and video.
---
Gemini isn't for everyone and everything, and that's kind of the point. It certainly doesn't seem like something meant for you, since you seem to be focused on research papers and apps. It's not trying to replace your OS, it's trying to be a part of it. Gemini also doesn't intend to replace the Web; it intends to focus on structured hypertext. The web became a steaming mess because of the feature overload you've described; the solution is to focus on being able to do fewer things and to restrict what's possible to prevent the same thing from happening.
I don't think we'll see eye-to-eye on this, because this looks like a value-based discussion on quality versus quantity to me when I don't think one is trying to replace the other. Alternatives aren't replacements; Gemini is an alternative, not a replacement.
Shameless plug for my own Gopher Client:
http://www.mattowen.co.uk/gopher/gopher-client-browser-for-w...
RSS in a terminal saves time and energy.
A voice UX was surprisingly easy to implement (given a good platform to build on), newer standards just get in the way of the experience!
I'll tackle a visual one soonish...
It’s been fun revisiting it as an adult!
[0]: https://www.brow.sh/
Gopher is a dead simple text-based protocol:
https://en.wikipedia.org/wiki/Gopher_(protocol)
Project Gemini is a slightly more powerful protocol for the small web:
https://gemini.circumlunar.space/
The tilde (~) community, which makes the small web social:
https://tilde.club/
I recommend you have a look at James Tomasino's Youtube channel, where he shows a bunch of cool stuff going on around Gemini, Gopher and tilde:
https://www.youtube.com/watch?v=DoEI6VzybDk
This is not for everyone, but if what you want is just basic text and functionality that can work from a terminal (though there are GUI browsers too), then this is perfect.