Ask HN: If you could reinvent the web browser, what would you do differently?

So, you feel the web should be more simplified? Or is it more that you want to keep the complexity, but also want good standards to be followed across all browser vendors (e.g. banish CSS prefixes and Chrome-only JS APIs etc)?

What about the wide range of web apps the new APIs enable? Do you think they should be catered for somehow?

I agree with this sentiment, but I'd take it a step further.

I personally think, the fact that we're shoehorning so much app-like stuff into the web highlights the need for a good web-based app delivery mechanism. If we provide a solid app delivery mechanism, then people can create their own document publishing apps/mechanisms using that.

I know this might lead to fragmentation of document formats, but hopefully there'll be one clear winner in terms of popularity and ubiquity (similar to WhatsApp or PDFs).

I'm actively exploring/implementing this, which has shaped my thoughts on the topic.

1. I'd have the browser recommend apps to install to open otherwise unsupported MIMEtypes & URI schemes. I think most desktops support this now (certainly FreeDesktop.Org-compatible Linux & BSD ones do via "AppStream"), it'll free webdevs from compatibility worries, & allow low-cohesian features like videoconferencing & (like brimstedt & osrec discussed) apps to be entirely independant codepaths.

2. For the sake of markettability & accessability (unlikely bedfellows) I like to target unusual human interface devices, e.g. voice, "smart" TVs. It should be a goal to work well absolutely anywhere!

3. I was kind of surprised to find that the next biggest simplification after dropping JavaScript is to move forms out-of-line from the webpage into their own codepaths. This allows me to better reuse OS code whilst simplifying hypertext event handling, avoid webdevs complaining that native widgets clash with their designs, & design a better UX for more devices.

4. There seems to be plenty of precedence suggesting it'd be a good idea to allow links to partially update a webpage, or open modal dialogs/popovers. hey.com is a prominant example.

5. The DOM needs to be replaced, it's an overengineered OO monster straight out of the 90's that can be largely replaced with about 10 lines of code in any language. It's expensive for the browser to construct, for JavaScript to optimize away, and for webdevs to deal with directly.

6. I've got almost no complaint against CSS (I found it easy enough to implement), but sometimes author styles actively make websites worse. So support for user & alternate styles should be emphasized to the user, and websites should still look good if they don't provide CSS.

7. Drop support for CSS `position:`, there's better alternatives now & it does more visual harm than good.

8. I prefer XHTML over HTML as it's easier to parse, but there's a better compromise to be reached between the two camps than the one WHATWG prescribes: handle self-closing tags and close any intermediate tags between the current one & the explicitly closed ones. This routine seems to work well enough on the modern Web, even if it'd break some CSS selectors.

That covers the main points, I'm curious to hear other thoughts?

Basically we should both maximize the potential for creative developers to create revolutionary tools, and also protect users who know next to nothing about problems with the web.

More vim-split like capabilities - native multiple tabs within a tab would be great imo

More programming languages (not just confined to Javascript)

More hackable (like emacs level hackable maybe)

The current plugins system is good, but maybe there are more creative ways to push the envelope.

Better native ABI support, maybe.. suppose I wanted to collaborate my Ableton jam session into someone else's Firefox tab, or I wanted to pair program in Sublime Text with somebody else via a Firefox plugin, etc

So kind of like how Netflix Party improves Netflix, but instead with any app

Maybe some good built in (opt in) tools that allow people who have a problem with surfing the net for too long every day. E.g. browsing screen time, etc. Of course private data, not to be shared.

Maybe a browser fingerprint-intercepting layer that just returns useless data to browser fingerprinting tools, unless the user would like to be fingerprinted. I really like Firefox's Facebook container plugin, which helps protect me from Facebook across the web.

Remove the JS runtime, replace it with a WASM runtime (assuming a future version of WASM with native, non JS DOM support) and relegate Javascript to a language module that just happens to ship with the browser for legacy reasons, along with other popular languages like Rust and Lua.

I would also re-introduce the type attribute to the HTML script tag and extend the tag to allow downloading WASM language runtimes as necessary. If the browser already has it cached, it would use the cached version like any other asset.

The end goal is for all software to be universally accessible and executable from a URL or some similar identifier. It's possible there may need to be another protocol for streaming applications rather than HTTP, I don't know.

This is a vaguely sketched out pie in the sky idea i've had for a while. I have no idea if it's even feasible, but I'd like to see it happen.