The great thing about something like an email service is that password guessing can be extremely rate limited. You miss three guesses and you can't log in for several hours. So an easily remembered password is perfectly fine unless it is blindingly obvious. As a homeless person loosing access to a phone on a regular basis, I am going to be comfortable with the risk that the Gmail password hashes might get leaked. I think others would be quite comfortable with that risk as well...
"There is an imperfect existing solution, with a problem, therefore we will ban the existing solution and move to a new, better one"
... should require extraordinary certainty in completeness of ones new solution before banning the previous.
There are very few times when the legacy method should be deprecated, and Google is the poster child of someone who shouldn't be trusted to recognize them.
(Looks pointedly at Chrome mv2/3 hubris and implementation clusterfuck)
I'm not sure why you think MV3 is a clusterfuck, it seems like it's doing exactly what Google wants. If you're confused by that, remember, you're the product, not the customer.
Pushing an implementation cutover by +6 months, and changing it from a hard to a soft date, because it has so many unresolved issues, incomplete APIs, and angry developers seems a fair definition of "clusterfuck."
Homeless people have no physically secure place to store their possessions. The reason so many of them lose cell phones is because they get stolen or destroyed. It's not because they're incapable of "keeping track" of them.
Just because 2FA "solves" the extremely narrowly defined problem, doesn't mean it is the best solution or even something that people can and will actually use. Upon those metrics alone, 2FA is usually one of the worst "solutions" to the problem.
Well, it isn't solving this one. Option to opt out would be nice.
> aren't capable of keeping track of a physical device for more than N weeks?
Bit ignorant of you. They could be just plainly stolen by someone else. A piece of rag working as a tent doesn't exactly have best physical security...
> I'm not unsympathetic to the problems of the homeless ant the burdens 2FA entails, but I'm also not willing to ignore the huge problems the 2FA solves, and realizing there will often be a tradeoff between making it very difficult to hack into accounts and making it easy for people with mental and other problems access their accounts.
It's not either or.
A homeless person has a vastly different cybersecurity paradigm, specifically, they don’t need much in the way of cybersecurity. Nobody is stealing a homeless person’s identity.
Given that, just let them disable it, and let them just use a password. It’s fine to rate limit them if they forget the password a few times, but let them keep trying to log in until they remember it.
It is clearly failing for this use case.
Security can't be seen as a one-size-fits all threat models. That will never be satisfactory, as requirements vary.
For most people in most scenarios 2FA is a net positive.
But denial of service is also a component of evaluating threat models. Here we're discussing cases where 2FA causes denial of service which is worse than any risk of getting the account stolen by password guessing.
Gosh, I don't know, how about literally all of the problems that 2FA solves in the first place? Passwords alone are a bad solution (often forgotten, easily re-used insecurely) for people without all of the challenges and frequent mental issues that accompany homelessness, why would you think they'd be a good solution for people who, as the OP says, aren't capable of keeping track of a physical device for more than N weeks?
I'm not unsympathetic to the problems of the homeless ant the burdens 2FA entails, but I'm also not willing to ignore the huge problems the 2FA solves, and realizing there will often be a tradeoff between making it very difficult to hack into accounts and making it easy for people with mental and other problems access their accounts.