Moderators don't have the capacity (and fairly, it is impossible) to check if they are bots or humans.
There are no good solutions, there are hundreds of thousands of intelligences out there, trained millions of hours on how to scam humans, capable of spitting out text tirelessly and shamelessly, and there will be only more of them, tens, hundreds, thousands times more.
https://news.ycombinator.com/newest - Scroll through there and there are a lot of [dead] submissions by green accounts. They aren't outright banned from submitting, but it often triggers auto moderation. It's like posting a link in one of your first few comments as a green user, that often results in shadow banning automatically.
I've always been learning by doing, in the course of making things, in order to keep making, with only a small fraction of looking at what others have done.
It's not just cookies, it's explicit consent to track you, and sell your browsing history to ~1500 spy companies around the world.
To the sibling comments: don't "accept the cookies" and then delete them.
- - -
I'm super angry at what the web has become, especially at the OS browser community. There is 0 browser (that I know of) that can access the web safely and conveniently. Atm I use Firefox with uBlock which blocks the cookie banners, but Firefox's extension model is broken, and every single extension provides 100% access to my websites to whoever controls the extension. I don't like it.
We need a browser with a safe extension model.
- - -
edit: I guess using 2 Firefox profiles, one with uBlock and one with my google/facebook/bank/amazon/etc accounts solves the threat posed by uBlock and extensions. I still don't like it.
Not just the web. Last time I installed Backdrops on my phone (a nice wallpaper app), you would literally approve hundreds of uses of your data when you press Consent. Even if you choose to manage choices, 200 'legitimate interest' options are enabled by default. Even when you are a paying Pro user. Data used includes location data.
What makes it worse is that a substantial portion of users block web trackers through an adblocker. However on phones, unless you have a rooted phone or use some DNS-based blocker, all these analytics get uploaded without restraint.
Atm I use Firefox with uBlock which blocks the cookie banners, but Firefox's extension model is broken, and every single extension provides 100% access to my websites to whoever controls the extension. I don't like it.
Some browsers (e.g. Vanadium, Vivaldi) have a built-in adblocker, so you have to trust one party less.
By that definition the majority of apps is malware. This was just one example, a lot of apps have an insane number of trackers. It was just to illustrate that even the simplest of apps do this.
> How would you implement ability to arbitrarily block any network connection on any website without giving an extension 100% access?
Browsers should provide a filtering option before they makes a request.
IMO a lot of no-brainer options are missing from personal computers. Like the ability to start a program with restricted access to files, network or OS calls (on Windows and on Linux). Browsers should provide the ability to inspect, and filter network access, run custom javascript on websites, etc.
We do sort of have that with the capabilities stuff (although I admit hardly anyone knows how to use it).
But the tricky part is that "reading files" is done all the time in ways you might not think of as "reading files". For example loading dynamic libraries involves reading files. Making network connections involves reading files (resolv.conf, hosts). Formatting text for a specific locale involves reading files. Working out the timezone involves reading files.
Even just echoing "hello" to the terminal involves reading files:
Capabilities are craaaazy coarse on Linux. Really only a small piece of the sandboxing puzzle. Flatpak, Bubblewrap, and Firejail each provide an overall fuller view of what sandboxing can be.
I like that uBO can manipulate the dom, so that I can remove annoying divs or animations that I don't want. Its so much more than blocking network requests which I can do with DNS anyway.
Safari’s extension model could be really good by now, had they not stopped putting effort into it. You are able to define which extensions have access to which websites, and if that applies always or only in non-Private¹ mode. You can also easily allow an extension access for one day on one website.
But there are couple of things I find subpar:
You can’t import/export a list of website permissions. For a couple of extensions I’d like to say “you have access to every website, except this narrow list” and be able to edit that list and share it between extensions.
On iOS, the only way to explicitly deny website access in an extension’s permissions is to first allow it, then change the configuration to deny. This is bonkers. As per the example above, to allow an extension access to everything except a narrow list of websites is to first allow access to all of them.
Finally, these permissions do not sync between macOS and iOS, which increases the maintenance burden.
What would a safe extension model look like to you?
At some point, you have to implicitly trust someone unless you audit every line of code (or write it yourself) and build everything from source that you run.
You need open source extensions (they are now, as the source is included) and you need to personally audit them, or you need to find a browser with every single feature you want.
Or do you want the browser to enforce permissions on extensions so you can lock them down as well as auditing them?
This is a solved problem for at least ad blockers for over a decade on iOS. The ad blocking extension gives Safari a list of URLs and regex expressions to block
No, it's a solved problem for ad blockers, a very specific problem case that extensions have traditionally solved. But the entire concept of extensions is far greater than just "ad blockers", although that's the use case for which 99.9% of people have used them for.
So you don’t “trust” Safari but you trust Firefox? In 25 years absolutely no one has accused Apple of storing your browsing data that’s not e2e encrypted (its stored so it can sync across devices).
I'm not the person who wants to redesign the browser extension ecosystem, but I can build Firefox from scratch and review the source code if I want, unlike Safari.
Once again, I'm not the one who said they would like to design a new browser extension framework, but I have created custom versions of Firefox that have all ability to phone home removed and modified extension support. So not verifying every single line of code, but making fairly substantial changes in the direction the parent poster wanted to go in.
I'm interested in a conversation about that, not you pestering me about whatever issue I seem to have triggered within you that resulted in your interjections in this conversation.
That the geeks solution to “I don’t trust $companyX” is that “I am going to compile an alternate solution without looking at the source code”. Is kind of meaningless.
How does "I can build Firefox from scratch and review the source code if I want" mean “I am going to compile an alternate solution without looking at the source code”?
They can always "access the network" in that the extension developer can push static updates for things like ad block lists or security updates.
It might be possible to have "read only" cross-tab access include automation APIs like keyboard + mouse, with user prompting to prevent data exfiltration.
That just seems like a lazy capitalism models. We had both 10 years ago without crazy tracking and accept all cookies why do we have for the worst lowest common denominator ?
I agree; the web ecosystem is enshittified garbage.
However, I'm just suggesting a modest improvement to browser extension security (that doesn't completely break ad blockers like Chrome's approach).
In practice, I run an ad blocker, and just trust that it won't exfiltrate bank passwords and stuff. Imagine the blast radius for a successful and undetected UBlock Origin supply chain attack!
My "pick one" approach (ad blockers would pick the middle option) would mean that comparable supply chain attacks would also need to include a sandbox zero day in the web browser.
I had similar frustrations and been maintaining a Firefox fork trying to fill a gap there. The result is Konform Browser and I think it might be relevant to you; please check it out!
> every single extension provides 100% access to my websites to whoever controls the extension
That feels a like a bit of overstatement and depends on what addons you use and how you install them... CSPs at least make it possible to restrict such things by policy (assuming user has been exposed to it and parsed it...). https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/Web... MV3 introduced further restrictions and controls regarding addon capabilities. While I agree the UI and UX around this could be much better, it's not all hopeless. The underlying pieces are mostly there.
While the fundamental addon execution security model in Konform Browser is inherited from upstream, for core addons like uBO you can improve the supply-chain security situation by loading it under "system scope" and disable addon updates in the browser itself. So while we don't (yet) improve on the runtime aspects you speak of, at least for now we can tighten up the supply-chain side to minimize risk of bad code running in the first place.
"Enterprise policy files" can be used to change Firefox behavior and tweak security model around addon loading. A little explanation and reference of how it works if you want to do the same in other FF build or for other addons: https://codeberg.org/konform-browser/source#bundled-extensio...
Any particular addon you think is missing from the list there and should also be packaged and easily available? Maybe will be able to improve some of the security-UI/UX here too down the line. I'd be keen to hear your take on how this should be done better!
Regarding what addons can and do leak about you to the outside... I think you may also take interest in FF Bug 1405971. We ship a patch for that which can hopefully be upstreamed Soon (tm).
Now that I think about it, if Claude can put most useful functions in a TUI and make them discoverable (show them in a list), than this could be better than asking for one-liners (and forgetting them) every single time.
That's weird because 2 years of one-to-one study with their parent should matter more than whatever method, also teachers don't just go and give up their method, even for good reasons. And 2 years of one-to-one study with their parent being effective is not a legit reason to change methods.
Moderators don't have the capacity (and fairly, it is impossible) to check if they are bots or humans.
There are no good solutions, there are hundreds of thousands of intelligences out there, trained millions of hours on how to scam humans, capable of spitting out text tirelessly and shamelessly, and there will be only more of them, tens, hundreds, thousands times more.
reply