Ask HN: What would you want in an ideal web browser?

Very lightweight, low CPU usage.

Extensible, easy to write addons.

Keyboard shortcuts, single-key shortcuts for everything.

A good bookmark manager.

Customisable till the very end, through a simple to understand menu instead of the about:config way.

Minimal, no mail or RSS client, just the vanilla browser.

Superb in-memory caching for instant page navigation.

In fact, the old Opera had nearly all these features before they went and messed it all up, making it a chrome clone in a different skin. I was a paid user since the 2.x days.

A few months ago I was thinking about the old MinWin improvements (to Windows' basic core services), how Windows 8.1/10 are (arguably) smaller and lighter than Vista, etc... and it clicked: Microsoft have realized they're no longer the "hardware industry killer app," if you will, and have moved out of the way... to let the Web take its place.

In my opinion, Web browsers are basically mind-bogglingly complex rube goldberg machines, based on conflicting standards that are also rube goldberg machines, and the whole scene is basically meta-rube-goldberg-ception.

A lot of the complexity and conflict is mostly due to the history and tradition behind the technologies we're using: on the one hand, all you want is arbitrary sandboxed code execution and sane UI; on the other hand, sites start delivering content exclusively via JS and everyone gets mad. And in the meantime, who audits their web browsers? What was that about "safe" code execution, where the browser protects you from "nasty" code? We hear about the browsers themselves doing the wrong thing by users almost every week.

We need to fight this. I've come to the conclusion that the only way this can be done is to simply make the effort and do the hard graft to port a bunch of existing applications to a new coherent, practical modality/paradigm, dump the whole thing on an unsuspecting world, and wait to see what happens next. It might catch on. It might not. Get feedback, try again. Keep trying until it goes viral enough to make a dent and improve people's lives.

There are so many interesting and fascinating UI experiments out there, but nobody seems to want to commit to actually practically trying out a few of them in real applications :( and I think that's what's holding everything back - everyone's waiting for everyone else to make the first move.

Here's where I'd start: everything should be able to send messages to everything else, to work around the language problem; everything should be an object with arbitrarily settable properties - everything from the files in the filesystem to the windows on the screen; with the appropriate security context surrounding everything, you could do crazy things like programmatically share the tags you attach to the files you share over P2P, or tag images with text descriptions of their contents; the console should be inherently textual but in such a way that it is exclusively touch-gesture-drivable; I strongly believe in the anti-mac/post-mac UX paradigm; I believe that it's impossible to create the "ultimate" solution that will work for everyone, and that all I can ever hope to do is nudge everything in what I think is a good general direction. And the only sane way to release an implementation like this is to put it into the public domain.

Agh, this probably reads like a rant, and I guess it is. I'm just not sure where to start, although I do have some ideas. Incidentally, I frequently use "asmqb7" on the web, including for my Gmail account.