There has to be a better way. This isn't working...
If Firefox & Chrome had support for something lightweight like Gemini (https://gemini.circumlunar.space/) then most sites could just use that.
With that sort of setup, restrictions on the web like uMatrix would be a lot less painful because most sites wouldn't ever need to be whitelisted.
But they wouldn't, because they couldn't track people.
They wouldn't use it (or the VAST majority wouldn't) because it means losing tracking.
If implementing your site as an app rather than using something like Gemini came at the price of making people think twice, in theory there might be an incentive to only do web apps when it's actually needed.
But of course not that many people would use that kind of discretion, not to mention "ooh, shiny!"
And it's hard to put the genie back in the bottle anyway.
A better web browser is really needed; one of things it can have is support for the Gemini protocol and file format. (It can also omit many things, as well as adding things.)
It might be nice to have a prompt like "This page wants to use a web font" like how you get for location. I realize that may be a lot of work to add in though.
Really, I'd just like to re-enable webfonts on Netflix so the subtitles don't look like garbage, but keep the fonts off on almost all websites. If anyone at Netflix is reading this, please fallback to something like sans-serif and not serif.
I wish servo was in a better place to fork
Prompts do help, but the prompt should allow the user to allow or block once or always, and to override settings too (e.g. for location, the user can optionally specify a location to use, or specify a command-line of a program that provides the location; for camera, the user can specify the path to the device or to a picture or video file, or a command-line of a program that will produce the picture). The user can then specify whether or not to always make this selection in the future, and in what scope. If the site uses TLS, or is a local file which has been digitally signed, then the user can optionally specify certificate pinning too, in which case the automatic selection will be ignored if the certificate does not match.
I also have other ideas, about meta-CSS, presentation mode, table of contents window, animation skipping, capability of loading animated GIF and PNG files as videos (so that you can rewind and pause it), better keyboard controls, save form data to local files, regular expression search, full cookie editor, SQL access to HTML tables, Xaw-style scrollbars, relative location bar, ARIA view, etc. Most of this is configured and/or activated only by the end user, not by the document. Although, some are capable in the document too, such as, a HTML document with a <video> referencing a GIF or PNG will work; such a file can be loaded with <img> or <video> and either way would work.
No "do not track" setting is needed. My idea is the user could set up request headers arbitrarily, as well as overriding response headers (including some "protected" ones, which are stripped if received from a remote site, so are only effective if set up by the user). For example, to enable "do not track", you can add the rule which adds the request header "DNT:1" with the criteria specified as "always".
I would also ensure that all timing APIs can be spoofed (including the JavaScript core Date object; in my opinion this is an I/O function so it shouldn't belong in the core). This is probably useful for testing, as well as for the end user to get rid of annoyances too.
In normal view and ARIA view, the viewport height would always be reported as Infinity. In print preview and in presentation view (which is a paged/screen media type), the viewport height is reported correctly.
Was it ever robust? Or just that there weren't yet so many exploits?
Qubes OS v4+ does not use typical software virtualization methods. VT-d hardware virtualization it uses was broken only once, and it was done by the Qubes founder: https://en.wikipedia.org/wiki/Blue_Pill_(software)