Upvoted because it's interesting that they're thinking like this, and relevant to HN. But it's still a horrible thought.

He says internet service providers (ISP) should have the power to sever internet connections if they detect a subscriber has infected computers

They have that potential now. They just need a service contract that allows them to do so, and it's as good as done. I can only interpret the fact that they're not doing so to mean that ISPs believe the cost of doing so exceeds the costs of dealing with spam, etc.

When an ISP detects that traffic flowing across its infrastructure bears the hallmark of a botnet - such as data being directed to a blacklisted address or sudden torrents of email traffic from a single machine - they would be empowered to quarantine their subscriber, destroy any malware found on the user's machine, and vaccinate it by installing the latest security software.

What a horrible idea. Cutting the user off is one thing. Forcibly applying security software is something else entirely.

Without a knowledge of how my system is set up, they're as likely to destroy it as to fix it.

For them to do this in the first place, they'd need root access to my system. I can't even begin to enumerate the ways this is bad.

And it's probably not effective anyway. If there's an official security software, or even a small list of them, it creates an easy target for the bad guys to work around. They'll build their attack code with the knowledge of what they need to avoid.