A different way to internet? Project Gemini

I’m interested if people have tried or have thoughts on this alternative to HTTP. There’s not really much content yet though.



I’ve used and it works quite well. Connected from cli to a cool site and it just works while being on a 2G connection tethered from a phone. For now though I don’t depend on any sites on this protocol. The very simplistic unix like approach it takes to everything is refreshing. I think it won’t be great in quite a few contexts but some it would be amazing in especially for those that want to access loads of information without the bloat of the modern web.

1 Like

I’m using it daily. There are some very nice capsules (sites) and people are coming with great ideas under such constraints. You can browse using the terminal, but I’m using Lagrange, a dedicated browser that I find very cool. I though the text-only philosophy would bore me quickly, but actually no, it’s indeed very refreshing and calm. I also absolutely love the protocol built-in certificate authentication.

1 Like

I like Gemini. Fast and light-weight. My favorite browsers for it are are Lagrange (GUI) and amfora (Terminal).

Both work well on the Pinephone,

This is what I noticed, it highlights the content-lite Las-Veges-strip of an attention grab that a huge amount of the internet has become.

1 Like

I totally get what they are trying to achieve, both as a reaction against bloated, unprivate HTTP and as a reaction against HTML and the associated bloat and spying, and also as a means to put more control back in the hands of the client device (and hence the user).

Will it take off? I have my doubts. While they say that they are not looking to replace the existing HTTP web as such, at the end of the day there is a vast amount of useful content on the existing web and a large investment in content and tools.

It remains to be seen how easy it would be to serve existing content out via Gemini (“dumbed down”) or to serve new Gemini content out via HTTP (probably not too difficult).

Have they gone too far with the simplification of “formatting”? Will people live without numbered lists? Without nested lists? Without tables? Without images legitimately embedded in pages? (as distinct from web bugs / excessive and unnecessary decoration and formatting / eye candy / emojis that go beyond Unicode) With very limited forms? Without subscripts and superscripts? Without inline links? …

Thought experiment: Take your favorite web site and ask yourself the question as to whether it could be done in Gemini and ask yourself what it would look like.

On the protocol side, the most obvious issue is the lack of “content length”. That means that you can’t pipeline multiple requests on the same connection. Granted that the simpler data format of Gemini means that you are less likely to need to send multiple requests without intervening user interaction (except when it is a robot doing indexing or archiving).

Combine that with mandatory TLS and that means maximizing the connection overhead. At the very least you would be relying on key reuse to abbreviate the TLS handshake.

The lack of any date (or other metadata) information means that caching isn’t really possible. The mitigating factor is that the lack of bloat means that content being transferred should be smaller anyway.

Yes and no. It’s good that they are “private/secure by default” (and in fact there is no alternative). On the other hand, it’s actually a barrier to entry, and the acceptance of self-signed certificates is a 50-50 design call.

In many situations encryption isn’t really needed. In other words you ideally want integrity but not confidentiality or privacy because the content is going to be public anyway and access to the content is not sensitive. TLS can already do that but there’s probably no way to specify when you want that and when you want full functionality. However that’s really a client implementation issue.

(Hey, I think I replied to the wrong post!)

Gemini is pretty cool. So’s i2p and ipfs.

I just installed amfora from debian bullseeye backports on one of my laptops.

I figured out how to navigate links, but when I tried to figure out the equivalent of “back arrow” in traditional gui browsers (I’ve never had trouble with the text based lynx). As I desperately tried key after key with seemingly no effect, I finally noticed that havoc was being wreaked on tabs in a gui browser on an attached monitor (amfora was running in a terminal window on the laptop display.

Back is ‘b’ :slight_smile: in Amfora.

As to it affecting the GUI browser, is it possible the GUI browser was active, and hence the keystrokes were going to it?

1 Like

Thanks, and I should have pressed ‘?’ early in my explorations rather than just now.

I use “focus follows mouse” policy, so typing in a text application in a terminal window should not send the focus to some other window and I get annoyed when that happens. I’m sometimes on a hair trigger about it because some applications don’t accept focus until I click (in the correct part of) the window.

This sounds insane to me. Why not just click to focus?

I did exactly that for years. Then around 25 years ago I came across “focus follows mouse”, liked it, never looked back, and become very frustrated and annoyed when it is not available or is implemented poorly.

My feeling is that it is insane to be forced to click to focus.

I tend to get the mouse cursor out of my way when I’m using the keyboard, so focus following the mouse would cause my workflow to descend into chaos. If you’ve got a handle on it then I can see the merit in using it. I myself don’t have the patience to retrain my brain for that.

I think they mean it will “replace” HTTP the same way that HTTP replaced GOPHER. There is still GOPHER content, and TELNET content for that matter, on the internet. Old protocols never die, they just fade away.