If I can have my "IDS" not only alert me but also protect the network by initiating proactive measures to block an attack I would rather have that over the simple alert notification.
Although some of the things being talked about here are a bit more advanced than what's been around, I don't think this is really anything revolutionary. Portsentry is one package I can think of that has been doing this for years now. I've never used it personally, but my understanding is that it can detect a port scan, which is usually the first stage of an attack, and then perform any number of responses ranging from logging the scanner's IP address to launching a retaliatory script of some sort. It can dynamically drop all packets originating from the attacker's IP and add the address to hosts.deny.

It would be interesting to see a product that could detect more advanced kinds of attacks than a portscan, though. I do agree that this is an area ripe for development, but it's not really a new idea.