There was one case in February (2005), which is a very, very difficult issue, and I'd rather not go into the specific product issue, but we were in violent disagreement, and for very good reasons. Actually, Larry and I were in agreement. Sergey was wrong, in my view. Sergey came in and he was just on a tear. He had convinced himself of this and he was absolutely sure. And I said, "Look, I can't take this, guys. You guys have got to sort this thing out. If the two of you can agree, I'll agree with you. But you have a deadline of tomorrow” Again here I am structuring it. "And you don't need to tell me the decision, you have to tell them the decision."

Tell who?

Tell the engineers. It's like 20 people who are completely demoralized because they're not getting clear direction. So what happens is the next night they get to the deadline -- and I remember it because this is an emotional issue -- and I called up Sergey, and I said, "What did you guys decide?" And he said, "I'm going there now to tell them," and he then described the solution, which was different than the three of us did -- and better.

So again the generic model is consensus building with dissent with a deadline. If you don't have dissent, stimulate the dissent, and inspect everything. That's sort of the default model, and then you manage the exceptions.

So the exceptions are when there really is a disagreement among the principles, and certainly I've encouraged people to say “I just don't agree,” and then every once in a while I've had to actually be a real CEO and mandate something and really force it and not listen to anyone else. Those are cases where it's legal or regulatory where I've just said, "Look, I'm just not going to participate in anything other than this outcome." And people know me well enough to know not to challenge me on it.