Ask HN: If you were rewriting Emacs from scratch, what would you do differently?

117 points by volemo ↗ HN
Don't get me wrong, I'm not planning on creating an Emacs killer, nor suggest anyone do that. But, hypothetically, what are some fundamental pitfalls of this foundational application?

339 comments

[ 4.1 ms ] story [ 256 ms ] thread
Not using Lisp would be helpful
Actually, I love Emacs for its Lisp! Yes, Emacs Lisp is not the best Lisp out there, however, IMHO, it's miles ahead of VimScript. If I were really to rewrite Emacs, I'd use some modern Scheme.
Fennel is a better Lisp than elisp. Neovim is extensible, and to a large degree written in, Lua, which the target language of Fennel.

Most developers do not like writing Lisp. That's just a fact, slamming the downvote button won't change it. I am not among those developers, I like writing Lisp, but most, flatly put, do not.

So by choosing Scheme you are competing with a remarkable number of little-used editors which can be extended in Scheme or Common Lisp, as well as Emacs, far and away the top dog in the extensible-in-Lisp-editor niche. Neovim has achieved the best of both worlds, because it can be extended in a rather nice Lisp, and also in Lua, which, while some find the quirks of the language annoying, is at least Algolic in structure, matching the mode of thinking and writing used by the vast majority of devs.

> Neovim has achieved the best of both worlds

No it has not. Fennel doesn't have the same level of integration into Neovim, like Elisp has in Emacs. Emacs is essentially a Lisp interpreter with a text editor built on top of it. This tight coupling allows Elisp to interact with and modify every aspect of Emacs, providing a level of customization and extensibility that is difficult to replicate with external languages.

While Neovim's approach with Lua and Fennel is commendable, it is unlikely that these languages will achieve the same level of seamless integration as Elisp within Emacs.

Then it wouldn't be Emacs. You're thinking of whatever, it could be a real nice picture in you head, whatever that you're thinking of, is not Emacs. Something like Emacs is not possible without Lisp.
- Use Common Lisp instead of Emacs Lisp.

- Design a more modular architecture to make it easier to extend and maintain different components.

- Design a more robust plugin system for development and distribution of extensions.

- Implement better sandboxing and security measures for extensions.

- Better APIs for extension developers.

- better multi-threading support baked into the editor.

> Use Common Lisp instead of Emacs Lisp.

Interesting, why not a scheme? Is it because the popularity in the industry? I don't know much about Emacs or lisps and looking to understand better

Scheme is smaller, has more static and less interactive philosophy. CL has most things you need straight out of the box. It's more like operating system than programming language. Exactly what Emacs wants to be.
elisp is more like CL than it is like Scheme; RMS was aware of CL when he was working on emacs, but was suspicious of some of its new features (like lexical binding).
Emacs' Elisp it's pretty close to Common Lisp, any CL user can learn Elisp in days (and the opposite it's true too). There's even a port of PAIP exercises into Elisp:

https://github.com/yfuna/paip-el

For the Common Lisp part, someone created lem [0], but for the other points I can't tell. I know there is an extension manager being developped, but am not able to judge the robustness. Also that there isn't org-mode.

[0] : https://github.com/lem-project/lem

> - Implement better sandboxing and security measures for extensions.

Why, pray tell, should this be of any concern at all? Sandboxing is used when the host is concerned about running programs that he doesn't trust. There is no reason that an Emacs package would require security measures around it, unless it were knowingly potentially malware. The only reality in which I could see this is if people were using proprietary Emacs extensions, in which case I would entirely understand it, because then people would be willingly running malware inside their editor. Perhaps this the stance VS Code users like to take towards extensions?

Well, Emacs does ship with a browser (because of course it does, that kind of thing is what makes Emacs so amazing) and we all remember the XZ Utils near-backdoor, so I think that security measures would be useful for people who decide to use a less trustworthy archive like MELPA or who install extensions with package-vc.
Just create another user account and test your new stuff there.
How does one just test for malicious code?
Oh, I think I see the strategy: just declare a problem trivial and then don't follow up in any way whatsoever. Genius!
The biggest security nightmare that web browsers have to deal with is that they are always continuously running random JavaScript. Emacs' browser doesn't handle JS, which is a good thing for a simple browser. Emacs doesn't share this problem in that the Elisp that you run is all chosen by you, and it's always free and never obfuscated.

I don't really see how the XZ backdoor is relevant to this conversation, since these types of exploits are the fear of server operators, and generally not people who run user programs like Emacs on their desktop. A rogue Elisp function could delete all of the files in your home directory if it wanted to, but people don't really complain about that because being able to delete files is a desirable function.

If you ask me, Emacs is a program where user freedom is and should be the ultimate goal. Adding security features that the user has to jump around 'for his own good' and only really serve as idiot nets are dubiously effective and only hinder this freedom. If someone wants to install a package from a source other than the official archives, then the trust is placed in him to actually look at the source code he's received. Even if you install an official package you should still look at the code, as it's quite useful to know what's going on inside it when you want to configure it or add functionality. The vast majority of packages I use are a single file and tend to not take up more than a few screens, so they are easily understandable.

A lot emacs usage today consists of running programs you shouldn't trust. And the rest of it is hard because you are either reviewing other people's libraries a lot, or simply not using other people's libraries, or updating infrequently.
> Sandboxing is used when the host is concerned about running programs that he doesn't trust.

Trust? Trusting criminals doesn't stop them from committing crime.

You may trust your emacs color theme author. Pretty colors from an innocent artist. You run the theme code without any sandboxing. Everything is going well. Then the author adds a keyloger, project code scrapper, and phone-home feature in his theme.

You update all your emacs packages automatically without any code review. Then you start getting emails from your companies security team asking why you uploaded sensitive projects to a 3rd party.

Wouldn't it make more sense to restirct color themes to color and font related tasks? Why should a color theme be allowed to scrape sensitive code from your disk and upload it to a 3rd party without your consent?

Read my other reply in this tree.

> You update all your emacs packages automatically without any code review. Then you start getting emails from your companies security team asking why you uploaded sensitive projects to a 3rd party.

If you do not trust the author or maintainers of a random program and refuse to review any code updates before installing it, then you are a moron.

I think if there is a concern that people would upload malicious packages, then there would be a level of trust put into the repositories that accept and offer them to review submissions before accepting them. This is still imperfect, but it shifts some responsibility off of you.

> Wouldn't it make more sense to restirct color themes to color and font related tasks? Why should a color theme be allowed to scrape sensitive code from your disk and upload it to a 3rd party without your consent?

Why SHOULDN'T a color theme be allowed to scrape code from your disk? Maybe the color theme is sophisticated enough to want to do that, or talk to some network. Something like a seasonal color theme that responds to the local weather might have a lot of hack value; and that is precisely the point you are missing. Emacs is not about restricting what you can and can't do, because the designers of Emacs understand that restricting what the user is allowed to do ultimately hinders freedom and creativity. It is one of the very few platforms left that's still like that, and I believe it should stay that way. If people want to use an editor that's very safe and tells them what to do rather than vice versa, then they should probably consider VS Code, or something like it which DOES upload your data to a third party without your consent, because it is smarter and knows better than you what you should be doing with your computer.

> If you do not trust...

Trust provides no protection.

> refuse to review any code updates before installing it, then you are a moron.

I personally do review every line of Emacs code I run. But I'd wager only a small handful of Emacs users do that.

> Why SHOULDN'T a color theme be allowed to scrape code from your disk?

Security.

> precisely the point you are missing.

Not missing it. nabla9's suggested security measures for extensions. You then asked why security is a concern. You seemed to imply there wasn't any reason for limiting extensions if you trust the author. That's simply not true, as you can easily be sabotaged by the most trusted color theme author.

> You seemed to imply there wasn't any reason for limiting extensions if you trust the author.

That may be implied but the precedent is not important. The bigger consequent point I am trying to make is that there is no good reason for limiting extensions at all, and my reasoning for that will follow.

> you can easily be sabotaged by the most trusted color theme author.

Okay. Let's say that we solved the hypothetical issue of colour theme authors sabotaging their themes by requiring color themes to use a format that can't evaluate arbitrary code, only dictate UI options.

Now what about all of the other packages that aren't color themes, that do more useful things and require greater breadth of functionality? I am not trying to separate things into functional and non-functional, as it is futile to do so--there are an infinite number of separations between the degrees of utility packages have, and once you have "solved" the issue of their insecurity by restricting one class of them, there are always more. And every time you do this you only degrade the freedom and extensibility of the entire system; you do not gain anything by it, you only lose.

This phenomena can be explained because it's really an application of Gödel's Incompleteness Theorem. For any program running on any computer, it will always be possible to induce conditions that the program was never designed to handle. Thus it is my belief that the entire field of computer security is really a joke, and anyone who sacrifices freedom for safety ultimately ends up with neither.

In practical terms a bit of security is useful, however Emacs is not the kind of program that benefits from it. It is not useful to waste time reasoning about which methods of security should be applied to a text editor. If you were using Emacs to run an air traffic control system then this type of thinking might make sense, but not as a reasonable or common case.

Support prettier typography (if the user is not interacting with Emacs through a terminal, in which case of course the typography is up to the terminal-emulation app). If text in Emacs looked as pretty as text on the web does, it would be less of a struggle for me to stay focused on the Emacs text. (Text on the web was already much above average in pleasantness to look at and to read in the 1990s.)

Get rid of any keybinding or UI convention that is there because that is the way they did it the AI Lab in 1967. Make the UI as familiar to the average computer user as possible (but keep the general design of a large rectangle of text) by using mainstream conventions (which come mainly from the Mac and Windows) for how to respond to this or that keypress or to clicking or dragging with this or that mouse button.

Inside Emacs is a cross-platform toolkit (where the platforms are MacOS, other Unix derivatives, Windows and the terminal) I would split Emacs into 2 projects: a toolkit and an app that uses the toolkit. That way, if someone wants to create an "standalone" org-mode app, Magit app or Gemini browser designed to appeal to people who do not want to spend any time learning to use Emacs the app or "Emacs the generalized interface to information", they have a straightforward way to do so. (These "standalone" apps that are as easy to learn as any other GUI app will I hope help popularize the Emacs ecosystem.)

One thing I definitely would not change is I would not make Emacs dependent on or closely integrated with a browser engine.

> If text in Emacs looked as pretty as text on the web does

Do you have an example of this? I can't tell any difference for the fonts that I use (with emacs-pgtk). I believe Emacs uses Harfbuzz (same as Chrom{e|ium}).

I have never figured out how to get a good font and rendering going for text in Urdu/arabic script.
Is vscode better at Urdu/arabic?
I have never tried but I assume you can do as well as a browser. But I would never use VSCode for writing generic notes. There is so much distracting mess in it.

I did setup Obsidian for a friend to write in Urdu, and that worked almost perfectly modulo some minor stuff.

Huh. I use pgtk Emacs, too, and am surprised to find someone who doesn't find my statement obvious.
Well, if there's no further info then I'm going to speculate you've misconfigured something ;)
Ok then, "make it harder to hold it wrong". :)
I'm not holding it wrong.
That's the joke, people weren't holding the iPhone 4 wrong, either. :)
Most of the text on the web for example is in a proportional-pitch typeface.

Does your Emacs usually use a proportional-pitch typeface? If so and you're on Linux, I'll install the font you are using.

I've tried using proportional typefaces in Emacs (on Mac), but there was something off, so I went back to monospaced. I could try again now that I have a Linux machine.

The text in my Emacs looks almost exactly like the text in my Gnome Terminal. (A slight difference in size is the only thing I notice. To be painfully precise, (window-system) evals to 'pgtk on my Emacs.)

The text in Gnome Terminal is not terrible, for sure, but text on the web is a nicer in my experience.

I used to use proportional pitch fonts for telega.el and certain document buffers, but I stopped because I find that with Jetbrains Mono (for me personally) there isn't any benefit even for longer text. I'd rather have everything be uniform.

Emacs is perfectly capable of rendering other fonts, too, though.

Yes, I remember now that having everything be uniform was one reason I stopped with the proportional faces in Emacs.
Pardon me if you're ahead of me on this, but it sounds like you might be using a proportional typeface as the default or fixed-pitch face. You should get nice-looking proportional type if you set the variable-pitch face to your desired typeface and enter variable-pitch-mode in the buffer. E.g.,

    (custom-set-faces '(variable-pitch ((t :family "Verdana" :height 180))))
And then in the buffer you wish to view with proportional type, M-x variable-pitch-mode.
Thanks for the info, but the result of that is not any better than the result of what I had already done. (I wrote a command that put an overlay on the buffer to change the typeface.)

It's not as good text on the web IMHO. Typography is very complicated, and I think the people who did the typographical details of Chrome and Firefox were very skilled, is my guess.

I use IBM Plex Sans (proportional) as my default Emacs font, with variable-pitch as a no-op (i.e., defined as "(variable-pitch ((t nil)))" in custom.el), and use IBM Plex Mono as fixed-pitch.

Tips for using a variable pitch font as the default:

0. Choose default fixed and variable pitch fonts with identical baseline-to-baseline heights for a given size; this makes everything described below work better (e.g., this is true for all fonts in the IBM Plex family across all platforms I regularly run GUI Emacs on [Linux, Mac, Windows]).

1. Define a fixed-pitch-mode by copy-pasting the built-in variable-pitch-mode and making the obvious changes (both are trivial applications of buffer-face-mode).

2. Add fixed-pitch-mode to hooks for modes that don't play nicely with variable-pitch fonts (calc, dired, hexl, magit, terminal and shell modes, etc.), or where you just prefer fixed-pitch modes (hint: define your fixed-pitch-mode in a package so you can use use-package's ":hook ((foo-mode bar-mode … baz-mode) . function)" syntax to manage this).

3. Some modes that pop up windows (frames in Emacs parlance) within editing buffers require extensions (e.g., company-posframe-mode for company-mode) to work properly in variable pitch buffers.

4. Last, but certainly not least: assign a convenient key binding to toggle fixed-pitch-mode. I can't emphasize this enough! In fact, I've found that variable pitch is fantastic for coding in most languages if and only if fixed pitch can be quickly toggled on and off with a keystroke, iff this setting is per file rather than global (and iff both fonts have identical line heights, but this is a feature of font families rather than editors).

For this reason alone, I'd argue that Emacs supports variable pitch fonts better than most text editors.

I use IBM Plex across the board and I was trying to understand the font issue because I do not have it. I default to fixed-pitch mode for everything, and use variable pitch for UI elements.
Instead of using ctrl and meta modifiers, use a leader key like escape or semicolon or comma or some such thing as the prefix key for key bindings. In fact, this desire for leader-key-based, non-modal text editing led me to write devil-mode for Emacs: <https://susam.github.io/devil/>.
Make CapsLock an additional Ctrl. On many old keyboards that is where the Ctrl key was positioned [1].

[1] https://en.m.wikipedia.org/wiki/Caps_Lock#Placement

(comment deleted)
I know many people like to remap Caps Lock to function as Ctrl. However that setup does not quite work for me. There is only one Caps Lock key on the left side of the keyboard. I need Ctrl on both sides of the keyboard, so that I can use the left Ctrl key while typing Ctrl+P but the right one while typing Ctrl+A.

There are other options as well, like remapping the Enter key to act as Ctrl when chorded or using sticky modifiers. I think using an ergonomic keyboard with two large Ctrl keys on both sides of the keyboard is probably the best solution. I've discussed some of these alternatives in more detail <https://susam.github.io/devil/#why>.

By the way, there are some vendors that still make Unix layout keyboards with the Ctrl key positioned where Caps Lock key usually is: <https://deskthority.net/wiki/Category:Keyboards_with_Unix_la...>.

I just leave Ctrl where it is and press it with the knuckle of my little finger. I do it just like the guy in the pic on this page, pressing the control key with his little finger: http://xahlee.info/kbd/how_to_press_control_key.html - because that pic is of my hand. (Xah might not recommend doing this all the time, but I've been doing this for nearly 20 years now and I've had no problems.)
For the past 3 years I’ve done the following remap Caps -> lctrl/esc, enter -> rctrl/enter, tab -> lalt/tab, backslash -> ralt/ backslash. Also have the physical right alt mapped to multi key/dead greek.

It works really well, and makes emacs much more comfortable to type in.

Downside is this keeps me locked to X11 and fighting the occasional app that reads the key codes directly.

Instead of making it an additional Ctrl key, you can also make it a new separate modifier key with XKB[1]. I've found this very useful over the years for WM-related key bindings, leaving the other modifiers for applications.

Tangentially, I really loathe how Wayland has no alternative to this. I'm expected to configure keyboard layouts in every DE or WM I use, which is a much worse UX.

[1]: https://vincent.bernat.ch/en/extending-xkb#attaching-symbols...

I was very disappointed that Apple gave up that fight.

They also at some point joined the PC world in moving the nubs on their keyboard from "d" and "k", where you were more likely to notice if your fingers are not in their proper place on the home row. Now if your right hand is offset slightly to the right, you won't feel anything, which is less immediately noticeable than if the nub were under the wrong finger.

I use a HHKB because of this. On the Acorn Archimedes I learnt to type on, it wasn't an extra Ctrl, it was the Ctrl.

Caps lock was rightfully relegated to bottom left.

emacs is not its keybindings. you can bind your emacs keyboard to do what you are asking for; as you said, you wrote a mode for emacs that works the way you want, and it wasn't necessary to rewrite Emacs.
> emacs is not its keybindings ... and it wasn't necessary to rewrite Emacs

That's indeed true! But the premise of this question explores the scenario: What if we did rewrite Emacs from scratch?

Making modal editing with a leader key the default rather than an option it currently is, is a pretty trivial thing to bring up as the stand-out "feature" a rewrite could provide.
Build it on top of Guile.
create an efficient async api for plugins. it’s kind of a kernel+apps situation. an extensible editor has to have a scripted plugin system, and they should be hosted by a core that is fast, can preempt plugins, and provides “ipc” btw plugins for advanced functionality.
The core should be in rust, verfied in lean, and the runtime should be in guile. Literate programming in org-mode should be a hard requirement. Package management should require patch algebra. Macros should be submitted to a leaderboard in a blockchain and yield flair in EUDC. M-x measure-beard-length should require 10000 hours of logged usage and unlock the major mode infinity-categorization.

Tongue-no-longer-in-cheek:

I reckon the C core of Emacs is some of the most battle-hardened code out there. Verification, a-la SEL4, is probably irrelevant but still nice. Guile is modern and performant but Elisp is still its own little joy. Literate programming is always nice until it gets in the way. Straight is good enough for me now. Macros are always cool and a leaderboard would be fun, but patch algebra is really nice, see jujutsu nowadays. And beard length is gendered and so only partially admissable. Infinity categories are way out there and always good for a reference.

Ok, you had me for the first three sentences.
Implemented mostly in Guile, with just some native primitives/kernel bits in Rust, makes a lot of sense, for a programmer's application platform in the spirit of Emacs.
> The core should be in rust, verfied in lean, and the runtime should be in guile. Literate programming in org-mode should be a hard requirement. Package management should require patch algebra.

Since I agreed with all of these, I’ll add some more non-ironic, idealistic wishes:

Plugins must be written in sandboxed WebAssembly so you can know what a plugin is capable of without reading the source code. The runtime must be portable so it can run in wasm32-wasi.

When I saw “should be in rust” my instinct was to flame like nobody’s business, glad I kept reading. Lmao. Good slow burn.

Though, not to put too fine a point on it, I know it was in jest, but the core implementation language is the least of my concerns. As long as it is extremely portable, compiles fast, and runs on virtually anything, then whatever the core is, doesn’t matter much to me.

Emacs is fine, but buggy as hell.

Their version of Lisp is clearly not suited for any large-scale development. (This trickles down hard into user experience, i.e., lack of parallelism or multithreading.)

> Emacs is fine, but buggy as hell.

Is this so obvious as to go without examples? I'm no Emacs power user, nor even really an Emacs user, but it certainly conflicts with my understanding of core Emacs.

I don't really know what a "core Emacs" is. It's all elisp scripts all the way down, and elisp is as brittle as bash for writing programs.

    Emacs is fine, but buggy as hell.
That kind of took 180 turn on that one.

Ship is fine, but leaks as hell.

As long as I can keep this ship, I will keep bailing.
Bugs aren't that big a deal for users usually though. Also all ships leak, that's not a big deal either.
Not using lisp
why the downvotes? This is a reasonable point of view.

I sometimes think using lisp for a language is a little like trying to implement comments within json data.

The downvotes are because "this reasonable point of view" is also an ignorant one.

Emacs just cannot be Emacs without Lisp. Emacs is a Lisp machine; it is tightly integrated with Lisp and needs Lisp to operate. Lisp is at its core principal model.

There are a number of different types of applications that are significantly more difficult to achieve with a non-homoiconic language; check out Hyperfiddle/Electric, there are some videos with demos. Why do you think there were so many attempts to re-create something like Org-mode, yet none of them were hugely successful?

For starters - the Lisp REPL has some significant differences; it just doesn't work the same way as a REPL in, e.g., Python. In Lisp, you can send any piece of code without any prelude or ceremony directly to the REPL. That REPL instance can even be on some remote computer, like a spacecraft millions of miles away (I'm not making up this shit, NASA did it at some point)

Anyway, Emacs is Emacs because of Lisp, not because someone on a whim decided to use Lisp instead of, I dunno, Pascal. Trying to build Emacs on top of some non-Lispy language is futile. It won't be Emacs. Sure, it might be to a certain degree even better, but it won't be even "like Emacs".

If I were rewriting Emacs, I would not replace its language with Python, Lua, or Cobol.

I don't like those languages, and I have just as much right to say that as the people who don't like Lisp.

- Can you not put any veggies in the salad?

- What? Okay, so fruits, berries?

- No, no, no fruits either.

- Hmm... what about croutons, nuts, maybe sunflower seeds?

- No, just use meat. I like meat. Just make it with a bunch of meat pieces, alright?

- Well, that is not a salad...

> Not using lisp

Don't you folks realize the imbecility of such statements? Emacs is a Lisp machine that allows writing and loading custom extensions based on Lisp! This level of extensibility and programmability is so deeply ingrained into Emacs, specifically because it revolves around Lisp. Emacs wouldn't be possible without Lisp.

Write Vim instead. /s
Emacs can perfectly emulate Vim already and even do more than Neovim ever could.
(comment deleted)
> If you were rewriting Emacs from scratch, what would you do differently?

UI: Electron, of course.

Json to represent the edit buffer in RAM. Each utf8 code point base64 encoded, in a json array, it itself, as a blob, base64 encoded. Now, before you complain that that is gonna blow up the data too much, don’t forget that 1. “Ram is cheap” and 2. “gzipped base64 is about the same size as binary”. So, of course, we’ll gzip the data in RAM.

Plugins should be JavaScript, as should be self-evident. And you’ll need a few installations of python (both 2 and 3) and node.js (each in its own docker container, obviously) to glue it all together and provide reproduceability.

With some care and work, it’ll run even on a modest machine taking up merely 60GB of disk, 32GB of RAM, a 4090ti GPU, and 8 CPU cores.

Every key press should be passed through an LLM, to add some intelligence to the editor. The user will, of course, supply a ChatGPT api key when they register for their mandatory myNewEmacs.ai account that they’ll need to subscribe to the editor for only the cost of a few lattes a month.

It is 2024, after all. One must use modern tools and technologies.

Prefer Tauri to Electron. It is 2024, after all.
Heh... The obvious choice of course is Unreal Engine. It's almost 2025, after all.
Are you sarcastically describing VSCode, I don't get it?
I think it's meant to be a sarcastic take of the next iteration after VS Code :)
I would change the development platform. Doing everything by mail makes things more difficult for people that are not used to the older mail+patch workflow. Having something like GitLab or sourcehut would be nice, as it would also bring a more modern bug tracker.

Personally I find following email conversations much harder than just a single conversation thread like in a GitHub issue, for example.

All sensible email clients have a 'thread view' for just this purpose, which effectively makes it a 'single conversation thread'.
Yes, the Gmail web client does this. However my (personal) problem comes more from reading the emacs-devel archives, where the thread view takes the shape of something more like a tree (maybe I'm not configuring something correctly). I was subscribed to emacs-devel at some point (which made reading easier) but it started filling up my account storage so I un-subscribed.
Use https://yhetil.org/emacs-devel to browse the threads. If you use mu4e or notmuch and often delete some older emails locally, but still want to read the whole thread, you can write some elisp helpers that would find the thread on point (based on email-id) or even download the whole thread.
I agree, and I've been observing this for years. It's now becoming a generational problem - younger programmers know and understand PR model, they don't want to deal with mailing threads and patches. These days, young people are like: "You sent me what? an email? Are you joking?" People shouldn't be catering for the comfort of the maintainers, no matter how arguably mailing threads are techologically more superior, it should be the opposite.
Don't use lisp. Normal people don't like it, it looks weird. I wonder how many projects failed because they were lisp. Normal people: Visits a project page. Sees it's a lisp. Closes page.
Parentheses scare away anyone who shouts "bro!" and fist-bumps each other, before they can insist "the first thing Emacs needs is a package manager, to hide code as much as possible from casual users" (missing half the point of Emacs).
Emacs has a package manager...
And consequently a lot more friction to users becoming extenders.
"A friction" you say? You can run Emacs, open a scratch buffer and extend it right away. You don't even have to save the damn code, you can try it out immediately. Folks complaining about Emacs being hard without even trying to understand any Lisp, is on the same level of whining about how web-development is so much harder compared to building shit in Squarespace (or something), only because you can't figure out HTML, CSS and Javascript.
Compare to when you'd go to use an add-on, and it would be one text file, right on your screen, and you would think:

. o O ( hey, I get some of this, and I can just start tweaking it here... )

You clearly don't know what you're talking about. You simply don't argue over customizability of Emacs, nobody does, because they know it's futile. If you think anything else has even some slightly better ergonomics to extend the thing, you just have not seen the bonkers level of extensibility what's possible with Emacs.

In Emacs, you can seamlessly integrate a function from a third-party package, say, a command that fetches a url, parses it, performs processing, and displays the results in a browser. Remarkably, you can modify it to send the results to an LLM or another function instead, without altering any other aspects. This level of granular control is complete bananas and only possible in Emacs. For VSCode, you'd likely need to create a new extension, while in Vim, you'd have to rewrite the entire function. Emacs, on the other hand, allows you to precisely specify and override only the desired part of the function. And once again, you don't even have to save a damn file to try it out.

So, yeah, I don't have to compare it with nothing. Nothing else comes even close.

I think I wasn't clear about what comparison I was talking about.

I've made some Emacs extensions, the public ones of which are at: https://www.neilvandyke.org/emacs/

My point was that I see a lot of people now who aren't getting the advantage that I had, of seeing "here's some Emacs Lisp code that does X", right up in their face, from the start, and constantly.

So they have more friction, to even knowing what Emacs Lisp looks like, and knowing how close they are to extending Emacs themselves.

I'm now utterly confused, even than before. From the start I thought you were saying "Emacs is hard to extend [for a newbie]", or something like that, and I've been arguing that it is not. Now I'm not sure what you're talking about at all - all the packages anyone uses come with their source code, the body of any function is a keystroke away.
I mean, it would be really nice if we didn’t have to drop into SVG or XWidget to mix text with GUI elements…

But agreed that Emacs is more user-extensible than all the other options presently out there, purely from how easy text-oriented extensions are.

Sure, yes, a graphical layer would be nice, I would love to be able to draw some arrows and other elements in some sort of an overlay, something like DrRacket does. And yes, better integration with an actual web browser would be splendid. I would love that.
Don't use Math. Normal people don't like it, it looks weird. I wonder how many projects failed because they wanted to use a lot of math notation. Normal people: Visits a project website. Sees it has equations and formulas. Closes page immediately.

That's how it sounds to me. Dismissing Lisp solely based on its syntax (that you're unfamiliar with), is equally irrational as rejecting projects that incorporate mathematical notation.

Allow for user controllable window layout within the frame, objectively the editor's only significant drawback.
god yes please. "Will I get a new window? If so, where will it be?"

If I could have Emacs with Acme-style window management, that'd be perfect

this is emacs territory so we have infinite customization via the `display-buffer` primitive.

see https://www.masteringemacs.org/article/demystifying-emacs-wi... for more information.

Thanks for reminding me about this, I really need to dig into it and get some rules in place.
unfortunately i cannot edit the previous post anymore. another _excellent_ resource is this article/video from "mr protesilaos stavrou" : https://protesilaos.com/codelog/2024-02-08-emacs-window-rule...

he goes by the moniker 'prot' and has tons of very very good emacs information (both on his website, and youtube). always worthwhile to check them out !

as someone would say, if you like that kind of thing, this the is kind of thing you will like :o)

After years of vim and windows parkour, I started using Emacs, and while the windows management was a bit irritating at first, I made my peace with it and use them in a more focused manner and rely more on the buffer list. I use register if I need a particular configuration.

Vim feels like working on a moodboard while Emacs is more a study desk.

I've been using Emacs for 25 years daily and this is still the thing that I think hasn't been really thought through. Yes there are hacks to mitigate that but there are still many rough edges with them.
What are you talking about? Emacs has incredibly good window control. It is in fact so good, people decided to build window managers on top of Emacs - see EXWM. Yes, the customization is complex and very confusing for beginners, but the levels of control are all there.
display-buffer-alist appeared only in version 24 and is still only a hack to bludgeon the editor into somewhat less insane condition than the default.
That was 12 years ago. Are you complaining that Emacs got sane window management "only" twelve years ago? Please don't ever get into Node.js ecosystem, the pace of accretion would make you cry, standing in the shower, rocking back and forth.
Buddy I started using Emacs at version 1.8, and no display-buffer-alist is not "sane". Just maintaining a stable window arrangement in the frame is not something easily possible.
I'm not sure what you exactly talking about, I'm having hard time separating "facts" from "personal opinion" here. I'm not sure what you mean by "stable window arrangement" to be honest. As a practical example I can only think of "jumping between places" feature, where If I, for example to have multiple windows and tabs open in the same Emacs frame (well technically speaking, tabs kinda make separate pseudo frames, but anyway), I currently don't have good mechanism of finding "the place", there's no good way to build "the breadcrumbs path" where you can traverse through all the places visited, Emacs afaik doesn't have good mechanism of finding exact tab/window/buffer/location, I don't know how much of it is related to your idea of "stable window arrangement".

Overall, I'm personally happy with Emacs window management, it's highly customizable and extensible, it gives you powerful keyboard-centric control, it supports complex layouts, it is deeply integrated within Emacs' vast ecosystem. There are packages like ace-window, winum, winner-mode, golden-ratio, shackle, popwin, eyebrowse, perspective, window-purpose, etc. I don't know what you're complaining about, I guess because I have not seen something even better than that.

Biggest issue w/ emacs is once you start adding packages everything is half-broken. Ideal emacs replacement would preserve the ease of extension and flexibility but have more of a static time verification systems (types and ...?) to prevent bugs (and bonus, make easier to achieve speed too).
I have been occasionally been hassled by Emacs package management for forty years.

A few years ago I very much shortened my .emacs file using the Straight package manager and I have been happier.

I would probably just implement vscode but for the terminal. Emacs shortcuts already work by default in vscode for the most part.
While Emacs is recognisable for its shortcuts, it is hardly a defining feature. Example: Doom Emacs adds Vim shortcuts, and it is still distinctly Emacs.

I think of VSCode as “Emacs, but JavaScript instead of Elisp.” That’s one thing I would not choose, in spite of the good things VSCode brings to the table.

Can I hit one key combination to edit the JavaScript corresponding to any vscode command, debug it and possibly modify it however I want? If not, it’s not really comparable to emacs IMO.
That's a fair point.

What I meant is that the runtime is arbitrarily extensible at runtime.

But I don't assume that people actually do what you describe.

VSCode is notepad with plugins, built upon a web engine. Comparing it to Emacs is a huge disservice to Emacs.
None of the main selling points of emacs have anything to do with its shortcuts or using it in the terminal. Plenty of emacs users (including me) rarely or never use it in the terminal. It’s a GUI editor just like vscode is.

I think this misconception comes from the fact that (1) people often compare emacs and vim, and (2) vim is usually used in the terminal. But emacs and vim are really categorically different things so I think the “emacs vs. vim” meme kinda doesn’t make sense.

Nothing
I too use vim.
Hmm, it's weird you forgot to tell everyone that you're also vegan and always recycle, and meditate daily. Guess what? Nobody cares...
Use python instead of lisp.
I agree.

Maybe it's just personal preference, since I think it's easier for me to think in python over lisp (which I've known for longer, but I still fumble through)

I do think python would make emacs more accessible to a wider audience.

> I do think python would make emacs more accessible

It would not be Emacs anymore. Emacs is specifically tied to Lisp. Emacs is not a text editor that uses Lisp as the configuration language. Emacs is a Lisp machine that has a text editor built into it.

An "Emacs-like" editor built on Python might be interesting, but it wouldn't be a "better Emacs" or even "like Emacs", it would be a completely different thing.

You may dislike Lisp, you may even hate Lisp, but the fact remains unchanged - there is an emerging class of applications that is significantly more difficult to build around non-homoiconic languages. Emacs is one of them. Stop fetishizing your favorite programming language as the quintessence of Emacs. The best one can do is to build a compiler/transpiler to spit out Lisp code, and people have tried that. Yet somehow, in over forty years nobody has succeeded in dethroning Elisp from ruling Emacs.

> Emacs is specifically tied to Lisp

GNU Emacs is tied to Elisp. But Emacs is a much wider family of editors written in a multitude of languages.

> Yet somehow, in over forty years nobody has succeeded in dethroning Elisp from ruling Emacs

That might have several reason. Maybe few people are interested to reimplement GNU Emacs in a different language. Like nobody has succeeded in dethroning C from the Linux kernel, using Lisp.

> Emacs is a much wider family of editors written in a multitude of languages.

Sure, there's Guile Emacs, there's MicroEmacs - both not Elisp-based, still built on top of Lisp dialects; there's XEmacs, Remacs - both still use Elisp, there's also mg which afaik completely not lisp-based, but I don't know how much of it still 'emacs-like'. In general though, GNU Emacs is what people usually mean when they speak about Emacs, unless they're explicitly talking about others.

If you say only Lisp can be used to implement Emacs, would you mean a specific Emacs or an editor belonging to the larger family of Emacs-like editors? When I would speak about how to implement "an Emacs", I would include the option to use different programming languages. For example I could imagine that C and Python is a valid combination, even Python alone would be an option. Python can be used interactively, which would be sufficient for an interactive extension language.

Here is an old Emacs timeline:

https://www.jwz.org/doc/emacs-timeline.html

The actual historic Emacs wasn't written in Lisp and was not extensible in Lisp. It was written on top of TECO and assembly. It was extensible.

The second and third Emacs were both written in Lisp (Zetalisp and Maclisp) and they were completely written in Lisp.

At some point Gosling wrote an Emacs in C and Mocklisp. Stallman took that one and rewrote it into GNU Emacs.

We have lots of Emacs variants written in languages like TECO, C, Fortran, ...

Craig A. Finseth wrote "The Craft of Text Editing: Emacs For The Modern World"

http://www.finseth.com/craft/

Chapter Ten of above book describes what Emacs-type means: http://www.finseth.com/craft/#c10

Extensibility is a general feature and not tied to Lisp or GNU Emacs.

The book contained a list of Emacs implementations. An newer list is here: http://www.finseth.com/emacs.html

You can see that there is a multitude of editors in the Emacs category. The list also mentions the implementation and the extension language.

> GNU Emacs is what people usually mean when they speak about Emacs

That's a bit sad. It's like saying "Linux" and think that its the same as "Debian Linux". Similar there are a lot of different Emacs-like editors. Claiming that there is only a single way to implement Emacs goes against the evidence that there are a lot of Emacs-like editors, which are not implemented in C + Emacs Lisp, including the original first Emacs.

I would think that by far the most important Emacs is GNU Emacs, but I don't think its implementation language choice (C + Lisp) is necessary to implement an extensible Emacs-like editor. Also be aware even though GNU Emacs is a popular Emacs editor, there are some people who are using different Emacs-like editors instead. I typically use a Hemlock variant written in Common Lisp and Zmacs, written in ZetaLisp. Both core designs date back many decades, actually even before GNU Emacs existed.

> I typically use a Hemlock variant written in Common Lisp and Zmacs, written in ZetaLisp.

So, I still can't see how that doesn't prove my point even further. These Emacs variants are based on Lisps, like you just said. I don't see anything "Emacs-like" today that's hugely based on a non-homoiconic language. Please, if you know any editor that allows me to modify the behavior of any given function/procedure/command with the same level of granularity as the advising mechanism of GNU Emacs, I would love to know about it.

Look at the list of Emacs-type editors. Plenty of them were not written in Lisp.

> Please, if you know any editor that allows me to modify the behavior of any given function/procedure/command with the same level of granularity as the advising mechanism of GNU Emacs, I would love to know about it.

Zmacs did that before GNU Emacs existed. It also allowed ALL parts of the editor to be changed, not just the ones written in Emacs Lisp for GNU Emacs. Remember, the core of GNU Emacs - both the core Lisp implementation and some core editor and UI functionality - is written in C.

Btw., on a real Lisp Machine the editor (Zmacs) was not the main user interface. For example on a Symbolics (but also in Interlisp-D), the listener and a file browsers were their own applications. Zmacs in Genera has a Dired mode, but no listener. Also Genera can run multiple Zmacs windows in the same Lisp, running concurrently -> the Lisp supports multiple threads and the applications use that feature.. Something which GNU Emacs can't easily do. It's mostly blocking and single threaded. Something which can't be easily fixed in Emacs Lisp.

Run Lisp code in GNU Emacs in the REPL (m-x ielm) and it blocks the UI. That can't trivially fixed and is a major implementation fail of GNU Emacs.

Other Emacs-like editors can run multiple things, without blocking the user interface.

> It also allowed ALL parts of the editor to be changed

Like interactively, without restarts and all? Wow.

> That can't trivially fixed and is a major implementation fail of GNU Emacs.

Yes, that is a real big, sad flop.

Dear Cooking Show, I really liked your recipe for the shrimp-avocado salad, and I tried to make it, but since I was out of avocados I just substituted them with potatoes. Also, it turns out I didn't have any shrimp, so I just used some hot dogs I found in my freezer. OMG, this recipe is amazing. You just made my day.

---

That's what using X instead of Lisp to make "a better Emacs" sounds like, okay? Emacs is a Lisp-machine, it's built on top of Lisp, it needs Lisp to be Emacs. Otherwise it wouldn't be Emacs. Like at all. Don't be stupid, stop saying stupid shit like "python instead of lisp"...

Don't be an ass.
What? I'm not trying to, I'm trying to be funny. Sorry if you read it in your head differently.
> Don't be stupid, stop saying stupid shit like "python instead of lisp".
Well, that is a stupid thing to say. You probably just haven't yet realized why.
You're just so full of yourself, aren't you? You know the one true way to do programming, it's lisp and everyone else is an idiot? This was old before you were even born.
Now, you're just being rude, I never said things you're implying I did, and I never claimed anything of being "one true way to do anything", I only tried to point out in a humorous way of your uninformed opinion, but I guess you lack basic sense of self-irony. Okay, if you rather remain sanctimonious, stiff-necked "professional", fine. I'm sorry if I hurt your tender feelings and I apologize if I angered you - I promise, my intentions were quite the opposite. Also, you have no idea when I was born, let's not get too personal, okay?
I do understand that for you Emacs is a Lisp development environment first, not a text editor.

Can you imagine people enjoying the user experience of Emacs while simultaneously preferring languages that are not Lisp?

> I do understand that for you Emacs is a Lisp development environment first

It's not "for me", it is what it is. Emacs first and foremost is a Lisp machine, with a text-editor built into it, not the other way around, it's not a text-editor that uses Lisp as its configuration language.

"user experience of Emacs" is to be able to send any expression and sub-expression without any preceding ceremony directly to the REPL. Everything what Emacs does stems from that. No, Python doesn't have the same REPL. Every single stage of it in Python is different. Lisp's Read, Eval, Print, Loop - they all have slightest differences. Those differences are possible because of Lisp's homoiconic nature. Because of that you get Lisp macros, because of that you get advising functions, because of that you can do source blocks in Org-mode that can interact with and modify their own execution environment. Lisp's homoiconicity allows code to be treated as data and vice versa, enabling powerful metaprogramming capabilities. This means you can:

1. Manipulate code at runtime

2. Create domain-specific languages easily

3. Extend the language itself

In Org-mode, this translates to source blocks that can:

1. Dynamically generate and execute code

2. Modify their own content or other blocks' content

3. Interact with the Emacs environment seamlessly

This level of flexibility and power isn't achievable in Python's REPL due to its more rigid separation between code and data. While Python is highly versatile, it lacks the deep introspection and self-modifying capabilities that Lisp's homoiconic nature provides, making Lisp uniquely suited for certain advanced metaprogramming tasks and interactive development paradigms.

So the bottom line is: you can imagine really hard, but it would remain just an imagination, if you want to build something like Emacs - a REPL that has a built-in text editor, you need a Lisp, because non-homoiconic languages DO NOT HAVE exactly same REPLs. Now, do you want me to get seriously pedantic and explain how every step in ReadEvalPrintLoop differs in Lisp and Python?

Between this and you calling me stupid - twice! I now understand enough about you to ignore you and feel good about that decision. Thank you.
I never called you stupid, I said the things you're saying sound stupid. Try to see the difference.
> if you want to build something like Emacs - a REPL that has a built-in text editor, you need a Lisp, because non-homoiconic languages DO NOT HAVE exactly same REPLs.

I don't think it makes sense. One can build programmable editors in many interactive languages. The language for an editor doesn't need to be Lisp. It could be Python, JavaScript, Ruby, PERL, Forth, ... Typically a form of EVAL or compile/load is enough to do so.

> One can build programmable editors in many interactive languages

Of course, one can - VSCode is a "programmable editor", no? Emacs is a bit different though, wouldn't you agree? It's rather a Lisp REPL that has a text-editor built on top of it.

A Lisp REPL is a user interface, not an implementation.

The first Emacs wasn't using Lisp at all. Wasn't it literally the Emacs?

> It's rather a Lisp REPL that has a text-editor built on top of it.

Just write an Emacs without a Lisp REPL. It can have the same Dired, just written in Python. There is nothing in Dired, which requires Lisp.

Dired is a file browser, sure, it can be written in anything. But what about something like Org-mode - with all its source-block magic, code execution, etc. etc.?
Why wouldn't one be able to implement this in JavaScript, Python, Smalltalk or even BASIC ?
GNU Emacs is a Lisp implementation with an abstraction layer over the hardware and OS. I would like to reserve "Lisp Machine" to computers with an actual Lisp operating system or emulations of those.
Of course. I understand that calling Emacs "a Lisp Machine" is quite a stretch, but due to the lack of any actual prominent hardware-based Lisp machines these days, I think it is a permissible simplification for the orange-site discussions. Yet it is, indeed, an important clarification, I appreciate it.
There are two open source emulators of actual Lisp Machines I would recommend:

Interlisp-D/Medley: https://interlisp.org

MIT CADR: https://tumbleweed.nu/lm-3/

These run an actual Lisp operating system.

These are interesting from historical educational point of interest, but they rather don't have practical use for modern software development, do they?
It's to keep the dream alive.
Scheme instead of ELISP. Concurrency (Async IO would do).
I'd make cursors be positioned within the document instead of on the screen. Currently, Emacs does not support off-screen cursors. If you attempt to scroll a cursor off screen, it will move within the document to stay on screen. This behavior is contrary to all modern text editors, and there is no good workaround. I once made a serious effort to start using Emacs, but ultimately stopped because of the annoying cursor behavior. (There were other annoyances, but none so fundamental and unfixable.)
I'll try to use the more common Windows/MacOS terms for it, but in emacs I often have the same file opened in two different panes within one window or two separate windows. I do this when I want to be looking at one part of a file while editing another.

Markers are used to mark a different point in the file and one can pop their location to a previous mark.

Putting cursors within the document does not preclude supporting different cursors for different windows.
This still annoys me slightly after nearly 20 years of using Emacs.

In response to keypresses, it doesn't bother me too much, as I'm used to the Windows-style behaviour of PgUp/PgDn/etc. moving the caret, much as the Mac behaviour of not doing that is sometimes useful. But for mouse wheel scrolling, which I do a lot - precisely because on Windows this typically does not move the caret! - having point follow along has never felt right.

> But for mouse wheel scrolling, which I do a lot - precisely because on Windows this typically does not move the caret! - having point follow along has never felt right.

Iam pretty sure you can customzize that if you want.

> scroll-preserve-screen-position is a variable defined in ‘C source code’.

> Its value is ‘keep’ > Original value was nil

> Controls if scroll commands move point to keep its screen position unchanged.

> A value of nil means point does not keep its screen position except > at the scroll margin or window boundary respectively.

> A value of t means point keeps its screen position if the scroll > command moved it vertically out of the window, e.g. when scrolling > by full screens. If point is within ‘next-screen-context-lines’ lines > from the edges of the window, point will typically not keep its screen > position when doing commands like ‘scroll-up-command’/‘scroll-down-command’ > and the like.

You’re misunderstanding. In those other apps the behavior is this: you open a document. You put the cursor on line 3 column 3. You use the mouse wheel or trackpad to scroll down to line 600. You hit cursor right. The cursor is now on line 3 column 4.

In emacs the cursor would be somewhere around line 600 instead.

It should be possible to save the cursor position, hide it if it moves away from view during scrolling, and restore it back on any command but scrolling, or at least at the firs cursor-movement command.
Conversely, I curse heavily at Mac programs where I can page down or scroll for a while, only to have an arrow press jump way the heck back in the document.
It (kinda) has multiple cursors per document if you split the frame and display the buffer twice. For longer source files I find myself splitting horizontally, editing in the left pane, and using the right pane for secondary movement and reference.

I have a buffer manager that can switch the displayed buffer quickly, when I switch to a different buffer and back again, it retains it's "secondary" cursor position that is separate from the other view.

(comment deleted)
This is a really interesting case of Emacs being so old that its default (and only in this case) behavior is just unable to comprehend current computer technology and usage. I think I am correct in saying that the lack of capabilities to have horizontal scrolling of a screen was an absolute hard limit of the teletype and terminal era. It should not be difficult to add such capability, but more than the obviously connected portions of Emacs’ code could implicitly rely on the assumptions that come from the impossibility of horizontally extending text.
This might be a dumb question, but how would an off-screen cursor work in a terminal? According to my understanding of the way terminal emulators/ncurses work, the cursor must be positioned somewhere in the screen, even if it isn't visible.
I'd be happy to break terminal mode (which I almost never use) in exchange for getting this feature.
whereas I use emacs exclusively in terminal mode, and would be rather upset to it no longer supported :)
Fair enough, but a straightforward solution would be to just not allow off-screen cursors in terminal mode, and allow them in GUI mode.
ncurses lets you move the cursor to arbitrary positions, and also lets you hide it. And you can disable echo, so you can type without the cursor having any influence on what you see. This means you can treat the terminal cursor as though it's a software-rendered cursor in a GUI app and apply all the same rules. But nobody considered doing this when Emacs was first written (maybe it wasn't possible then), so the assumption that the cursor is always on screen is difficult to change.
It is possible to emulate this in Emacs Lisp. I've created a package to that extent, see http://ankarstrom.se/~john/etc/emacs/scroll-without-point.el. (I've made a few changes to it the last couple of weeks that I have yet to upload, though.) I should probably publish it somewhere else eventually, but I've always felt it to be a bit of a hack. Still, I use it and it works well 99% of the time.
Curious, why would it matter that the cursor stays where you were when paging down?
The exact behavior isn't hugely important (although I think off-screen cursors are a useful feature). What matters is consistency between applications. There's no realistic scenario where I do everything in Emacs, so Emacs has to behave in the same way as other applications. If not, I risk data loss or corruption because I might fail to notice the cursor ending up somewhere I didn't expect. This is especially likely to happen with a feature like off-screen cursors that I use only occasionally.
When you scroll away you can jump back by hitting anything on your keyboard. Also you can stay where you are, by clicking anywhere. It is a useful feature, that enables you to scroll to your heart's content.
You might know, but you can set marks to go back when you scroll off. I have to admit it's not as convenient as the cursor just behaving like that automatically, though.
Let us write plugins in whatever programming language we want and provide some simple interface like Unix sockets or something for us to communicate with the editor process.

Instead of making the first class installable extensions plugins create a primitive called modes that encapsulate groups of plugins with extra configurations so that instead of having to pick every plugins for our setup we just pick the most popular javascript or ruby etc. mode and then add a couple of our own plugins on top.

Add some system that suggests hotkeys based on usage. If I hit l 20 times instead of just hitting f’ to get to the end of the line show a popup. Gamify the key maps and suggest key maps and features even from my plugins.

Instead of having a package manager for your editor just use homebrew and integrate it deeply into the editor.

Just use Acme. TCL, Perl, Go, awk... on sockets/files, a la Unix.
I don't care that much about the implementation, but I would like to see an emacs environment with:

- instant startup

- blazingly fast scrolling

- minimal keypress-to-display latency

I have written a lot of elisp (and had to deal with with buffer variables, dynamic scope, etc.), but aligning with modern scheme or lisp (if it can be kept compact and efficient) probably makes sense at this point. (Current emacs should provide backward compatibility as needed.)

Since so many people use emacs as an IDE, I think having an official emacs co-project/subproject focusing on a standard, extensible IDE framework (perhaps for other apps as well) would make sense.

It still seems like a pain to display graphics in emacs in various terminal apps. This should be easy, fast, and standardized.

As others have noted, supply chain attacks against open source are rampant, so vetting and sandboxing packages seems to be more important now.

I'm confused. On the modern devices I've recently used emacs on (including very low-powered raspberry pi devices), all of your three criteria are already true.

What kind of HW are you running emacs on where this isn't the case?

Emacs pauses are almost always due to some operation blocking in the main thread. It's pretty annoying and is mostly a consequence to a lot of things effectively being single-threaded. Stock emacs doesn't do this very often, but third party packages that might do expensive tasks often do
Ah. A case of "don't do that, then".
a) Build it with native-compilation flag

b) Use https://github.com/blahgeek/emacs-lsp-booster

c) Tweak GC vars, see: https://emacs-lsp.github.io/lsp-mode/page/performance/

d) Use plists for deserialization (described in the previous link)

e) Learn how to use built-in profiler and don't hesitate to launch it, sometimes simply disabling some minor-mode in specific setting is all it takes

f) Current state of Tree-sitter in Emacs is a mixed bag - for some modes you get huge improvements, for others it's the opposite. YMMV.

This is from 2015, but emacs still has higher typing latency than vim in 2024.

https://pavelfatin.com/typing-with-pleasure/

If you have better numbers and comparisons for emacs keypress-to-pixel latency, etc. I'd be interested.

Note classic vi started up faster than vim.

Device makes little to no difference: Raspberry Pi 5, MacBook Pro M1, ThinkPad P1, etc. - emacs is clunky on all of them. I like emacs and it's my daily driver. But it isn't fast.

see also:

"Computer latency: 1997-2017" https://danluu.com/input-lag/

"Measuring keyboard-to-photon latency with a light sensor" https://thume.ca/2020/05/20/making-a-latency-tester/

I have often see some fellow coworkers start up their fairly complex vim/neovim/other-vim setups and quite frankly… they ended up recreating a poor emacs implementation.

Yeah it’s faster at starting up but it’s worse in pretty much everything else.

Oh and they end up typing Ctrl-this Ctrl-that anyway.

Might as well use gnu emacs.

Receiving output from a process, as is done for example by shell mode and (I guess, but have not verified) compilation mode, is slow in Emacs because all output is run through comint-carriage-motion and ansi-color-filter-region (which does nothing but throw color information away IIRC) which are written in Elisp.

If you don't run the output through those 2 functions, then non-printing characters remain in the output that severely undermine legibility.

C programmers tend to develop the ability to glean information from the voluminous output of build processes as it goes whizzing by on a terminal, so they find compilation mode frustratingly slow. Or at least that is my guess as to what happens.

For instant startup I recommend daemon mode and emacsclient. Start emacs via emacs --daemon, open windows (in the terminal or otherwise) via `emacsclient` / `emacsclient -nc` (I use aliases for these).
Yep, warm start is the only way to go unfortunately. But I want a fast start from cold boot like classic vi.
I think running daemon mode has significant advantages. You can have frames running on multiple virtual desktops for different uses. You can also step away from a buffer, and return to it, maybe from some other location on a different computer.
Though trying to run Emacs like vi is a pitfall in itself. Clunking around the terminal summoning Emacs to edit files is really not a particularly effective way to use it. Emacs is not vi, and using Emacs like Emacs makes more sense. Most users explore your filesystem and spawn and monitor processes from within Emacs, rather than from a terminal.

Not that you can't work that way (and using `emacsclient` is the way to do it), but you're losing out by trying to shoe-horn usage patterns familiar from other toolchains onto Emacs. Emacs is not a text editor, it's a text manipulation platform that includes an editor - a number of editors in fact, including a vim clone.

I wrap my slow .emacs loads inside #'run-with-idle-timer so they run when I'm not looking, and not likely to notice.
what is this point of this, when would you not be using emacs?
Only needs a few seconds of idle time to load libraries.
The solution to "this text editor takes too long to start up" is poorly addressed by the solution to "just leave it running in the background all the time".
It really is not poorly addressed, as it's a reasonable solution to the issue which also brings about it's own benefits. If you have your system configured such that the Emacs server starts up when you log in, by the time you actually start Emacs it will long have already loaded completely. If you actually use your Emacs for several things throughout the day, including but not limited to reading and editing text and documents, then it is likely you will have it always running during that time with several files open, and will find it useful to be using Emacs in the same state it was in the last time you opened it.

As an aside, having it running in the background was actually the way the original EMACS that ran on ITS was intended to be used: When you C-x C-c out of it, it does not kill the program, it simply puts it in the background, and the next time you invoke EMACS it brings it back up. This is similar in effect to hitting C-z, then using 'fg' to bring it up again, but this only really works in a terminal. The Emacs server achieves the intended effect better, especially on a graphical display.

It's not a "solution"; you still need to solve your undesirably slow startup, but if you're having to kill and restart Emacs multiple times a day, something is wrong with your workflow. Normally, Emacs doesn't require restart - for weeks and months in some cases.

emacsclient is useful on its own. For example, I use it to edit just about any text, in any input box in any app, delegating the task to Emacs, or when I want to open some URL in Emacs, while browsing a web page in my browser, or to send some selected text to Emacs.

This has strong "you're holding it wrong" energy.
Well, if you see someone launching a browser, visiting website, then quitting it, and then later launching it again to visit another webpage, you'd tell them it ain't right, no?
> instant startup

My config is Doom-based. I use about 300 different third-party packages, and I don't even know how many built-in ones. It takes about 1.3s to start. Wishing for it to start even faster is like wanting my microwave to warm my tortilla in 5 seconds instead of 30. I don't restart Emacs every day, nor do I eat tortillas daily. Use a modern package-manager, and defer the loading of packages, it will start very fast - just like 'emacs -q' does.

> minimal keypress-to-display latency

A lot of times, setting keyboard rate is all it takes to make it nice,

On Linux: xset r rate 170 80

On Mac:

defaults write NSGlobalDomain KeyRepeat -int 1

defaults write NSGlobalDomain InitialKeyRepeat -int 10

I'd start with the core TECO editor I've written in Free Pascal[1].

Free Pascal does gigabyte strings you don't even have to allocate. Then I'd read the Emacs manual, and start writing code and tweaking TECO to make it all impedance match better.

But I'm old and weird, so maybe not the best way to get there.

[1] https://github.com/mikewarot/teco