I use firefox, and find this function useful:

     # Open a new firefox instance, with a temporary profile
     function firefox_temp() {
        dir=$(mktemp -d)
        echo "Profile directory: ${dir}"
        firefox --new-instance --profile ${dir}
        echo "Cleaning up .."
        rm -rf ${dir}
     }

I'm not familiar with Chrome, but perhaps there is a similar "profile-directory" argument you can point to a temporary location?

You can use the `--user-data-dir` arg to specify a profile location when opening chrome. So, on a mac, something like:

  open -Fna /Applications/Google\ Chrome.app \
  --args --user-data-dir=/my_temp_profile_dir/my_tmp_profile

If you copy an existing profile to `my_tmp_profile` chrome will use that, otherwise it behaves like a new setup.

That's interesting... does it behave much differently to private windows?

If we assume that the private-mode is "secure", then I suspect there's probably not a huge difference in practice.

But this way I'm very certain extensions, history, local-storage, etc, are absolutely not available.

(To be honest I tend to use this mostly avoid caching, and other things. Useful when you're testing redirects and new designs.)

There is a flaw in your proposal: Deinstalling doesn’t necessarily ensure all program data is removed (at least on desktop OSes). When you reinstall the program, your previous program data can still exist and be used as is. So the persistent cookies can actually be super persistent in that they survive the program removal.

Some people run Chrome in a container:

https://github.com/jessfraz/dockerfiles/blob/master/chrome/s...

It could be configured to run in a totally clean container to keep the browser reasonably isolated from the host system.

Of course, the truly harcore can use Qubes OS and spawn a disposable VM to browse the web from:

https://www.qubes-os.org/doc/disposablevm/

That lead me to https://blog.jessfraz.com/post/docker-containers-on-the-desk... Nice!

IIRC Chrome installs some kind of persistent daemon that is entirely separate from the application itself. Keystone Agent or something like that.

I've come across this before. I don't know what its full responsibilities are, but I know it's part of the Chrome update process.

I came across it when I was trying to get two totally separate Chrome installs working on one machine (using https://github.com/djbclark/chrome-schismator), in such a way that they can run in parallel, update in parallel, each totally isolated from one another. It seems as though it's not possible to have two Chrome installs update themselves separately due to the Keystone application only being able to store one reference to a Chrome install at a time, meaning that the other Chrome instance never detects that it's out of date.

Perhaps you should use Firefox.

Perhaps Firefox should finally get around to not being a destroyer of performance, stability and battery life on a large subset of MacBooks, so people such as myself can use it.

I honestly never had that problem, sure performance wasn't great 5 or 6 years ago. I've used Firefox as my primary browser for at least five years, switching from Chrome when performance started to improve for Firefox.

Agreed. When I open up Brave on my phone to use Google Newsreel or something, I feel that it is a bit more snappy than Firefox. Firefox also crashes more often than I would like on my ArchLinux personal laptop. However, Firefox rendering speed has caught up with Chrome since the Quantum update, maybe the rest will catch up also.

I have Firefox (with accidentally permanently 50+ tabs open) on a touch bar MBP and it's fine.

It doesn't seem to affect everyone/every machine.

Then it doesn't seem fair to blanket blame FF on macOS. A few more details would go a long way to legitimize the complaint.

If a user tests Chrome and Firefox and only Firefox has bad performance/battery life, they'll blame Firefox.

Mozilla is aware - there are a few bugzilla reports open about this - but like most Firefox bugs it takes months or years to be fixed. Last year (nov/dec?) they fixed something that improved battery life, but it's still worse than Chrome for me.

Mozilla has been tracking cpu utilization/power consumption issues with certain MacBooks since Quantum was released. It seems to mostly affect MacBook Retinas without discrete GPUs. From the big tracker it appears the latest theory is it caused by an interaction between Firefox's compositor and the macOS compositor.

Nah, they rather implement a file sharing website.

Just thinking. Maybe using lxd with rendering on the X server of the host if you use Linux. Then keeping a base container that you can clone anytime you want to start a new browsing session. Or making use snapshots if you like.

Browsers should be on our side. If they're failing at that, then they're not good.

Personally, I'd like browsers to be much smarter about which cookies to keep and which to throw away. Keep cookies about settings, throw away everything else. And instead of not accepting cookies, leading to sites complaining about that, just accept them and throw them away at the end of the session.

Also, I'd really like the option to open some sites always in incognito mode. Many sites have no business knowing who I even am, but some really do. Same cookie policy for all sites makes no sense. Cookies need to be a lot more transparent, and instead of relying on annoying and incomprehensible GDPR popups, the browser should take care of that.

You can use docker for this