Before Chrome's initial beta launch, we had constrained popup windows instead of a popup blocker. IIRC, the way they worked was:
1. Popups would be constrained to the renderer viewport (i.e. script couldn't move them outside those bounds).
2. The popups would behave as part of the tab. If another tab got focused, the popup would hide with the tab that popped it up.
3. If the tab is dragged around, the popup would drag with it.
4. *Users* could drag the popups out of the constrained range, at which point they would behave like popup windows do today.

Popups are a source of much frustration for users today, but they are also the only way to achieve some uses cases:
1. Popping out a chat to a new window
2. oauth or payment systems that need assurance that they're not occluded by other page content.

Constrained popups fix the user frustration by minimizing the need for a user to manage them, e.g. they can close them by closing the tab that opened them, they always know which tab opened the popup, etc.

The reason they didn't work in the past is because we tried to use them as a replacement for popup blocking instead of something that's complementary. Users were seeing too many popups.