That's because patio11 is telling you how to keep your site alive under load, while that page is telling you how to decrease your page load times. They're contrary goals in this case. Keep-Alive makes things load faster, but it puts a cap on how many clients can connect to your server before it curls up and dies. It would be ideal if Apache would let you set a high-water limit for Keep-Alive connections after which it turns the feature off, but I don't know any way to do that. You can set how long they're kept alive, and you can set how many requests are allowed per Keep-Alive session, but not how many sessions are kept alive.
This is only true for servers that use separate thread/process for each request... It doesn't apply to event-driven servers (nginx, etc).
I'd even say that keep-alive is always your friend and the longer you can keep connection open the better... Of course there are always OS-level limits (open file descriptors, etc), so you should use LRU-queue on idle keep-alived connections to make sure that you won't run out of resources...
The best way to handle keepalives is to proxy everything through nginx, which can handle a huge number of connections with very little memory. Turn keepalives on in nginx, turn them off on your app server.
