Gemini is a new internet protocol | Hacker News Reader

Gemini is a new internet protocol

admin.flounder.online

67 pointsmodinfo72 comments

Main problem of Gemini is that it's read-only, by design.

The author of the protocol didn't want to add forms etc, as that adds complexity; but as a result, there is no way to write things. There are no discussions or a way to create "pods" without going back to some other protocol.

I think as a result, most content of Gemini is talking about how Gemini is great and nothing much else.

So we need to move the transaction management for the connection out of the document and into a conversation.

A transaction looks like an email conversation, with server response containing enough metadata to generate a new request to some endpoint that can process it.

The next request to the Gemini page will need to have some GUID to locate the conversation and obtain the latest version.

This sounds like something that might not be too insane over a VPN, where one owns the whole conversation.

Otherwise, it could be fraught with peril.

what you cite as a bug feels like a refreshing feature to me. Looking at how the Lagrange [1] client interprets Gemini shows how rich and enjoyable Gemini can be, and it leaves me optimistic that we could browse it in the future on something like an ereader. The constraints that it imposes makes it so the author is encouraged to write text, not a website.

This doesn't have to reproduce the entire suite of features that a modern web application offers.

[1] https://gmi.skyjake.fi/lagrange/

Aren't e-readers perfectly able to display e.g. MD over HTTP on existing clients instead of gemtext over Gemini on a one off client? Or even gemtext without it being over Gemini or a dedicated client? Similarly how does being incapable of doing something enhance just not doing that something?

I get Gemini as a way to form a clique of like minded people who know of it but I don't get Gemini as enabling anything that couldn't or wasn't already being done before.

Gemini takes a while to appreciate.

The first step is realizing that it was deliberately designed to be inextensible. Anything that can be extended will accrue bloat.

The second step is to realize that it's really just plain text. Once you realize that, the choices about what kinds of formatting make total sense:

   * bulleted lists, but not hierarchical
   * links but they have to be in a paragraph of their own
   * preformatted blocks but no italics
   * linewrapping but no collapsing of multiple sequential linebreaks
   * headings but no boldface
   * blockquote with > just like email (*are you listening, @dang?*)

They're the kinds of formatting that impose zero burden on somebody reading through a terminal. In fact, linewrapping is really the only thing you need to do to gemtext to make it readable (versus just "cat foo.gemini"). Rich clients like Lagrange then take this limited set and render it as beautifully as possible on something much more powerful than a VT100.

In a way, this is just an extension of the "first step". If you start allowing rich markup, where do you stop? Not allowing anything that would make the markup itself any less readable provides a natural point at which to halt extensibility, for principled reasons.

I recently discovered that Lagrange will treat a link to an image as a lazily-loaded inline image -- you have to click the link to make the image appear, but it appears inline where the link had been instead of replacing the page that linked to it. I think this is a really beautiful way to enhance the experience for people reading through a GUI without tempting authors to do anything that would burden people who are reading through a terminal.

The only big wart on the whole thing is TLS. It's cute that it solves the TCP truncation problem, but that's a lot of bloat to pay in return. I guess there isn't yet something at Layer 4 that is as minimal, elegant, and battle-tested as wireguard (Layer 3) to take the place of TLS though.

>They're the kinds of formatting that impose zero burden on somebody reading through a terminal.

Interesting. From a modern Web POV a lot of these are backwards, but if you look at it from terminal and other usage it is quite refreshing.

You are correct, ereaders have supported viewing the web for years, but it has been a miserable experience since the beginning.

    * Inline links everywhere tempting you to click away to a new page, both fracturing your attention as well as being a multiple-second delay between pages
    * Images are everywhere, needing to be interpreted by the rendering engine of the ereader either with a simple monochromatic threshold to figure out what parts of the fully color image are black and white, or if you're lucky it has some way to re-process the image through dithering so that it's possible to see what it is
    * web layouts always wanting you to have a screen larger than an ereader can support, so you're stuck scrolling back and forth on a tiny screen that already has a miserable refresh rate of once or twice a second.
    * the web is an application platform now, the vast majority of its features just not fitting within what an ereader can support

You're all correct, you can easily implement what Gemini does if you make one website that is only text. In fact, you can probably do it better since you don't have to be constrained to gemtext, or constrained to its handful of weekend-project servers for hosting the site, all of which have their own issues.

But the thing that keeps drawing me back to is is that while you can easily implement Gemini with HTML and HTTP, it is impossible to implement HTML and HTTP in Gemini. The fact that you can't implement a webapp in Gemini means that if you're browsing pages in Gemini, you get a consistent experience and every site is clean and respectful of the experience (whether the author wanted to respect it or not, they have no choice).

Perhaps the real counter-argument to Gemini is not "why don't website authors just make their websites simpler?" but instead "why don't modern browsers run in 'reader mode' by default?".

Where did I say anything about websites or HTML? Since when can HTTP and HTML be implemented in Gemini? And even if it could how does that answer the question?

For all the explanation of what Gemini does and why HTML+HTTP in a browser don't this response does very little to actually answer the question: "Aren't e-readers perfectly able to display e.g. MD over HTTP on existing clients instead of gemtext over Gemini on a one off client?"

Lagrange is very good. It seems I can't set a custom homepage URL, I hope he adds that at some point.

Can anyone recommend a good de facto homepage as a jumping off point for exploring what's out there on Gemini today? There are a few decent directory sites but in particular it would be nice to find a page with links to new or updated sites.

I use Antenna:

=> gemini://warmedal.se/~antenna/

Note: anyone can submit an RSS/Atom feed to Antenna, but submissions from Techrights.org and Kiwi Farms are blocked by default.

At the moment, you can set one or more homepages via bookmarks: right-click on a bookmark and select "Use as Homepage".

That's... not really true. The spec clearly allows for user input.

Section 3.2.1: https://gemini.circumlunar.space/docs/specification.html

Yes, for a simple, singular query, like a search. It's not meant to be a longer (or more structured) input. There are "message boards" that hack around it and try to make it a text input, and it feels like that, a hack.

The solution to "Gemini doesn't do X" is "make a client that does X".

For example: a combination Gemini/IRC client would be pretty neat. Gemini documents can include IRC discussion links and open right up in the same client.

A lot of the early HTTP-based web was read-only - I mean how many Geocities pages were there in the mid-90's with nothing more interactive than emails and maybe little things like a guest book or whatever.

Agreed. In general, composing together simple standards seems a far better option than trying to organically expand e.g. web standards to handle everything under the sun.

Gemini is a little gem - https://news.ycombinator.com/item?id=30072085 - Jan 2022 (122 comments)

Gemini is Solutionism - https://news.ycombinator.com/item?id=30067400 - Jan 2022 (218 comments)

Lagrange: A desktop GUI client for Gemini - https://news.ycombinator.com/item?id=29291392 - Nov 2021 (90 comments)

Gemini: The Misaligned Incentives - https://news.ycombinator.com/item?id=28688232 - Sept 2021 (84 comments)

What is this Gemini thing, and why am I excited about it? (2020) - https://news.ycombinator.com/item?id=28600436 - Sept 2021 (208 comments)

Gemini's "uselessness" is its killer feature - https://news.ycombinator.com/item?id=27490769 - June 2021 (193 comments)

Gemini Space - https://news.ycombinator.com/item?id=26670464 - April 2021 (27 comments)

Agate, a simple Gemini server written in Rust - https://news.ycombinator.com/item?id=26401158 - March 2021 (34 comments)

gemini:// space - https://news.ycombinator.com/item?id=25986378 - Feb 2021 (170 comments)

The Tragedy of Gemini - https://news.ycombinator.com/item?id=25807633 - Jan 2021 (28 comments)

Hacker News over Gemini - https://news.ycombinator.com/item?id=25225810 - Nov 2020 (21 comments)

A Gopher View of Gemini - https://news.ycombinator.com/item?id=25005307 - Nov 2020 (9 comments)

A look at the Gemini protocol: a brutally simple alternative to the web - https://news.ycombinator.com/item?id=23730408 - July 2020 (347 comments)

Castor: A browser for the small internet (Gemini, Gopher, Finger) - https://news.ycombinator.com/item?id=23161922 - May 2020 (75 comments)

Gemini – A new, collaboratively designed internet protocol - https://news.ycombinator.com/item?id=23042424 - May 2020 (62 comments)

I've been enjoying Gemini using Lagrange on desktop and Deedum on mobile. It's reminiscent of the early web: the kind of people who put up Gemini capsules are not talking about the kind of stuff you see on every popular website, so you get detailed idiosyncratic takes on history, technology, politics, religion, crafts — all manner of things. For example, I'm ardently opposed to a particular political point of view, and using Gemini I've found a lot of interesting, introspective writing by people who adhere to it that has given me a new, if not more positive, perspective that I could not have gained on the more public web.

It's worth checking out.

Can you suggest any links? I’d be curious to read some of the political discussions you mentioned.

Just go to gemini://geminispace.info/ and start searching. You'll find more interesting things on your own.

I find it interesting and admirable when people can strongly disagree but still have some empathy for opposing views. Any chance you could share the subject matter of the viewpoints you reference? Maybe even describing the points of view without stating which view you’re ardently against?

I'm a conservative libertarian, reading what communists have to say for themselves.

Thank you for that♥

Personally I got tired of it. Almost everyone is of the same leftist, punk, ecological mindset. If you're into diverse groups, Gemini isn't the place. I don't feel like I'm learning anything new from Gemini beyond the first few weeks of spending time there. Maybe if you don't have many leftist friends you'll find it more novel, but I found it full of the same talking points and little else.

I also find the place way too judgemental. The community loves to hate on anyone that doesn't share their aesthetic.

> Almost everyone [on Gemini] is of the same leftist, punk, ecological mindset.

I definitively am (my capsule is gemini://idiomdrottning.org and I write about all three of those things) but I often feel kind of alone with that perspective. There's a lot of "let's go Branden" type curmudgeonliness which I'm not into, or weird Time Cube type stuff. I only know of a handful of other leftist capsules (love them ♥).

Also, While the merits of left-wing policies and of punk aesthetics can certainly be called into question, criticizing "ecological" is like saying "and the posters on there are subject to gravity! They fall downwards, and they have the gall to live their lives going forwards in time!" We all live on the same Earth and wrecking it affects all of us.

> Maybe if you don't have many leftist friends you'll find it more novel, but I found it full of the same talking points and little else.

It's not my intention to re-harp on the same points already made a thousand times but to instead afford a new way of thinking about systems and markets. I try to update older posts rather than make new ones about the same thing.

> I also find the place way too judgemental. The community loves to hate on anyone that doesn't share their aesthetic.

Here, I can only agree, and I'm often guilty of this, too. I see how people wanna warp or distort the aesthetic (for example, produce spec-breaking sites or accessibility obstacles) but as I try to clarify it I end up inadvertently slamming those who aren't into the whole minimalism brain trip and that's a real problem.

> There's a lot of "let's go Branden" type curmudgeonliness

:( That's unfortunate. I find the leftist stuff a bit repetitive, but certainly more aware than that sort of curmudgeonliness. I guess it's a natural consequence of the growing population of Gemini huh.

> criticizing "ecological" is like saying "and the posters on there are subject to gravity! They fall downwards, and they have the gall to live their lives going forwards in time!" We all live on the same Earth and wrecking it affects all of us

Oh I'm not criticizing the ideology at all. Ecology drives a lot of my personal politics as well. And specifically by "ecological" I mean "ecological punk" (messed up with the comma in my original post) here which I'm referring to as a low-tech/minimalist approach to ecological sustainability. I just find it overrepresented and a bit repetitive on Gemini. I can find myself nodding and agreeing, but few of the posts challenge me to think and grow the way talking with individuals with very different ideas of ecology (say more maximalist approaches) would. At the end of the day, for me, living ecologically sustainably is the goal, not _how_ we get there.

Thanks for your patience with how I misread your comment. I appreciate that a lot, and I agree fully with what you're saying re the limits of solar punk and minimalism compared to more complex system solutions.

I understand the impulse to simplify things, but there’s an important question that needs to be addressed before doing so. It is the following: why won’t this new, simple protocol accumulate the same complexity as it’s predecessor over time?

The Gemini guys actually think about this a lot so they're working to make the protocol the least extensible as possible or not extensible at all. They plan to freeze the spec forever after 1.0

That doesn't stop anything. There is nothing stopping me from writing up a spec that says "Put an annotation like this in your gemini file and with tools supporting this spec addition, you will see an image embedded". For non compliant browsers you will just see plain text and for new extended browsers you will see an inline image.

I agree that this is possible and that the Gemini project might be futile because of this.

Sounds like they want to guarantee this never actually catches on. Pretty sure they will succeed.

Because no one actually uses Gemini other than to write smug posts about how text only is the best.

One thing: a version number is explicitly not supported by the protocol or standard. There is only one Gemini version now and forever.

There is realistically no version numbers on the modern web. Its a living standard where new features come in whenever. The underlying communications spec is versioned but this has little to do with possible features and only needs to be updated for performance and security.

That's true for HTML, but not HTTP. HTML is organic - here's your document, good luck making it work. HTTP is clearly versioned with significantly different changing features at each revision.

You don't need to touch HTTP to extend features though. Most of what people care about is the HTML/CSS/JS web api which can extend at any time. No user or developer was ever excited about a new version of HTTP which really only helps performance.

Oh hey, I wrote this. It's pretty out of date, but it reflected the ideas that motivated me to get into Gemini in the first place. Hope it encourages you all to get involved too!

I agree with other commenters that Gemini isn’t really needed so much as a culture of web minimalism. One amazing and perhaps revolutionary thing about the web is that everything still works. If we wanted to, we could just go back to table layouts and minimal CSS. It’s not like vendor software/“apps” which no longer support, eg Word 2004.

Part of what draws people to Gemini is not just web minimalism in terms of architecture, but minimalism in terms of culture. The web is no longer the exclusive niche of a technically literate elite, but in terms of the "geeks, MOPS and sociopaths" model of subcultural evolution[0], Gemini is meant to be a space where the MOPS and sociopaths never show up to ruin things.

Once you understand that, it makes sense. It's the tech version of "my own web with blackjack and hookers and no Homers allowed."

Except not with blackjack and hookers, because blackjack would require complexity and interactivity and hookers would mean capitalism and advertising. But definitely no Homers.

[0]https://meaningness.com/geeks-mops-sociopaths

Interesting, thanks for this.

I like some of the ideas behind Gemini, but I'll never get onboard with something made by an individual and not actually standardized by a reputable standards body. BDFL type software will only end up not being accessible to some people and not thinking about concerns that aren't important to the one or two people controlling it (even if a larger open source community develops around it; it's still largely controlled by the person who set the initial direction and did most of the work initially with no collaboration or cooperation from others).

Could you provide an example of a standards body along the lines of what you would accept which is not primarily comprised of and funded by commercial interests?

This is a bit of a straw man; many (most?) standards bodies aren't driven by commercial interest. I agree that there are some that are (3gp, probably wc3 though I'm not super familiar with them I just hear people complain about that as a problem all the time), but many if not most are almost entirely volunteer or have a more mixed way of doing things (ie. the within the IETF plenty of companies put their people in the bigger working groups, but also many volunteers who aren't affiliated with a company driving the work are also a part of them or completely control them). Others take an absolutely-no-company approach like the XSF who only allows three people who work for the same company to join and requires all members to be acting of their own accord and not on behalf of a company. Sure, some standards bodies have a problem with commercial interests, but it's not the big problem your question implies.

Getting to the point of being stamped by a reputable standards body is something that comes after experimenting in public for a while in a non standard way

Gemini has already won. Gemini users have carved out a space for a thriving community, which doesn't need to scale to be valuable and fun.

I don't like the idea of re-inventing Internet Standards.

Maybe we can create a culture around creating content using as little web features as possible? I really like minimalism but I also really like interoperability and improving upon existing standards.

The problem I've noticed is you can't get enough people to respect the minimalism. Yes you could implement Gemini in HTTP by having everyone play by the rules, but how many of the links even on HN go to pages with ads, autoplay videos, and "sign up for updates" modal popups?

When I browse Gemini with Lagrange ( https://gmi.skyjake.fi/lagrange/ ) if you avoid orange links, then you're guaranteed to have a quiet, text-only experience.

Sounds like you want AMP...

Ah! I wrote a really long-winded sarcastic joke response only to see you lay it out so explicitly :p

Maybe we can build on top of HTML but only allow a restricted subset. Something that will allow pages to load fast and avoid obnoxious advertising. How could we get people to use this subset vs. pulling in other parts of the web we don’t like? We could create a validator for that subset and make certain web properties exclusive to content which passes that validator.

Cat-V proposed HTTP 0.2 (<http://http02.cat-v.org/>) as a subset of HTTP and I think the idea is neat. Tho I also personally don't like that HTTP gets blamed for the modern web's shortcoming when I think it's the one part that seems to work really well.

What I find interesting is the same argument of simplicity and that an Engineer can implement the protocol in an afternoon are the same talking points for the Gopher protocol. I’m mildly interested in Gopher and curious why Gemini over Gopher?

In the document at https://gemini.circumlunar.space/docs/faq.gmi Ctrl-F for "shortcomings" and you'll see "2.2 Which shortcomings of Gopher does Gemini overcome?" where they make their case.

Sounds like a pretty good case for enhancing gopher instead of starting over.

I made a whole video about this. Why gemini over gopher and why gemini over http -- https://www.youtube.com/watch?v=9Q3GCzG2gvE

> While these were initially relatively simple protocols, today, they are enormously complex, to the point that a web browser like Google Chrome has upwards of 5 million lines of code.

How many of those lines of code are dedicated to HTTP and HTML vs all the other things browsers do? How many of those lines of code are for CSS, JS, UI, Extensions, security, local storage, etc?

I just downloaded a copy of the 1-page version of the latest W3C HTML standard document. This document is 11.65 megabytes and over 80,000 lines long. It is safe to say that HTML alone comprises a daunting level of complexity.

Gemini mostly seems to be comparable to HTTP, since the markup/formatting part seems to be extremely minimal.

For HTTP, if we narrow things down to HTTP/1.1 GET without keepalive (connection-reuse) and for fixed content-length documents only, then I don't think it's vastly more complicated than Gemini. A client are not required to implement to implement more if they don't need it.

Browsers implement more (HTTP/[23], other request types, caching, etc) because it improves the user experience - but it wouldn't even be required for showing most web pages.

I understand and empathize with the desire to restore the Internet as it previously existed as a constellation of weird neighborhoods and niche hangouts. Gemini feels like a technological solution to a social/cultural problem though.

I used Gopher and Veronica back in their heyday. The Internet was a fantastically cool novelty back in the days of bag phones and exorbitantly expensive long-distance calling, and I totally get wanting to recapture that feeling. I think it's a lost cause, though. Large parts of the world suddenly becoming easily and mostly instantly connected was a sort of species-level watershed moment. Those of us that were around for those early years were lucky to get to see it, but I don't think it can ever really be recreated, except as some sort of LARP.

What does this blog article really do that the official page doesn't?

>Writing Gemini text takes maybe five minutes to learn

So does learning HTML if you only want Gemini level formatting. There is also nothing preventing you from just serving plain text files over HTTP either.

I decided to package up a fully portable version (including Downloads path and registry keys) for Windows: https://portableapps.com/apps/internet/lagrange-portable

If Gemini pods could be more easily deployed (think Vercel), I could see it being more used for personal blogs.

How can we solve the chicken and egg problem here? Is there a way we can bootstrap the use of Gemini off of the main web?

This proprietary protocol controlled by a single entity offers nothing over Gopher.

I don't understand how anyone would give it the light of day.

How is Gemini a "proprietary" protocol? The protocol specification is entirely out there in the open. [0]

[0] https://gemini.circumlunar.space/docs/specification.gmi

I never said closed. I said proprietary.

Relative to software licenses, this would be the sort where the source code is there and you can read it, but that is all i.e. not Open Source nor Free Software.

It is not specified by a standards board or the like. It is an immutable protocol made by somebody's whim.

And it isn't very good. An example: It will never have a header with the size of the payload. The only way you will find out is by downloading the whole payload.

Okay, if we go with proprietary meaning strictly sole developer, and ignore that it has been developed with a community, first by mailing list, and now on Usenet [0]...

Gemini was first published roughly in 2019. At about three years old, what protocol has a standards board? It's too young.

It took C roughly 17 years to get a standards body.

It took Gopher three years to get the first RFC [1]. Was it a proprietary protocol? It was the same age as Gemini currently is.

[0] https://web.archive.org/web/20211026220037/https://lists.orb...

[1] https://datatracker.ietf.org/doc/html/rfc1436

>At about three years old, what protocol has a standards board? It's too young.

Your point is moot due to Gemini's immutability.

It isn't a protocol carefully designed by a committee. It is just somebody's brain fart. It cannot be fixed and we are giving it too much attention.

> It isn't a protocol carefully designed by a committee

https://en.wikipedia.org/wiki/Design_by_committee

Standardized specifications are usually expensive to read, making them inaccessible to hobbyist developers. The Gemini specification is free.

Reply on news.ycombinator.com