Comment by munificent | Hacker News Reader

Comment by munificent | Hacker News Reader

Parent Full thread

Caveat: I work at Google but know nothing about this area and my opinion here is entirely personal.

> which would create an unsustainable situation at this scale.

Financial sustainability may have something to do with it, but I suspect the larger issue is that providing too much detail essentially trains malware authors to route around the company's defenses.

Imagine the Play Store as a castle which has both good townsfolk coming and going as well as being perpetually under siege by a malicious lord. Sometimes, the castle's defenses inadvertently prevent a townsperson from getting to market to sell their onions. When the townsperson is like, "Hey, I can't get in to sell my onions." it's helpful for the castle defenses to be like, "Well, we have the portcullis raised from 9am-11am on Tuesdays and the gatekeepers listen for your accent to decide if you're a local or an enemy."

But that's, like, exactly not what you want to say if the "townsperson" you're talking to is actually an enemy spy taking notes.

Say it with me now:

>"Rough consensus, and running code. We are not the Protocol Police."

Half the problems we have nowadays is because we have manufacturers playing "the Program Police", which leads inevitably to the point you just made.

You are now, like it or not, adversarial to any User looking to do anything you find unconformant with your bottom line. You cannot solve these issues by whitelisting, just like you can't solve the problem of crime by whitelisting, and hiding the conformance suite. If you can't know the test, you can spend infinite cycles changing the wrong thing to comply with it, and I do not find that to be a tenable state-of-affairs to push on users, even if intentionally aimed at the malicious ones. This is the same problem we have in meatspace with our overly byzantine legal system; but nobody accepts that secret laws are a good idea because if everyone can read the law, it's a national security risk. At least no one without some serious conflicts of interest.

Do you really think that your company is going to nail down a good solution to a problem that society at large can't even handle reasonably? I mean, think about it. This really is a subset of the general question of how to keep everybody doing something productive. I don't even need an answer. I just want to encourage people to think.

> >"Rough consensus, and running code. We are not the Protocol Police."

This model absolutely does not work when it comes to creating spaces where humans interact. There are bad actors and someone has to police them or they will abuse other users.

If you run a bar, you have to hire bouncers. It's simply part of the cost of hosting a safe venue.

I suspect the larger issue is that providing too much detail essentially trains malware authors to route around the company's defenses.

Perhaps so, but it seems not unreasonable to have SOME ability to work with the creator of an app that's been on the store for years with a substantial number of ongoing users and (speculating) a non troublesome patten of installs and purchases.

Nobody believes that Google is technically out financially unable to do this, which leaves the other option - at a corporate level not giving a shit enough to even bother trying.

Google will often do the right thing whether by plan or by happenstance, but it pays to be aware that when it does the wrong thing there is no recourse and will be no correction.

I'm sorry, but the "security" excuse is BS. You don't have to tell users what automated tool flagged them or how their violation was discovered.

You do have an ethical obligation to inform them of what policy was violated with sufficient detail that a good actor has a reasonable chance of complying with your policy.

I think that this should be required of any company that to provides publicly available goods/services, not just Google. This doesn't just help with monopolies, but also makes it harder to hide racism and censorship behind opaque policies.

That doesn't seem to be a problem in this case? Telling spammers they are blocked due to copyrighted images trains them not to upload copyrighted images. Win-win.

picking up copyrighted images is another indicator that user X is a spammer, providing that info would eliminate the signal

Well, this is the essence of discrimination and we wouldn't tolerate it for a whole range of indicators (you're black, gay, if a particular race, etc etc). My guess is the real reason they won't tell people is that they would end up in court pretty quick.

so, in you mind, detecting copyrighted images and using that as a metric to detect spammers is discrimination? Are antivirus programs discriminating too??

I bet you indent your code in an inclusive way

From a definitional point of view yes. Using an attribute to place someone in a class and then making decisions on a class basis without actual evidence they possess the other attributes of the class is discriminatory behavior.

Reply on news.ycombinator.com