Cisco has developed two ways to manage a group of access switches in a somewhat consolidated fashion.
The first on the scene was switch clustering. This provided the means to manage a group of switches utilizing only one management IP address. It did not provide any sort of fault tolerance or centralized configuration (you had to "hop" from the master switch to the others and still configure them individually). Most network engineers felt the benefits of this approach did not out weigh the negatives and it was seldom deployed outside environments where CiscoWorks/LMS was also deployed.
The second approach was switch stacking. This required that the switches provided the ability to be connected utilizing a "stacking cable" that provided a shared control and possibly data plane. It provides centralized management of all switches (i.e. you can configure them all from one switch) as well was fault tolerance as any stack member is a potential master and can take over if the master fails.
Since you mention 2960G switches, AFAIK none of them had the option to utilize switch stacking so you are only left with clustering. I would recommend not clustering them unless you really have a need and the ability to properly manage them. It probably won't provide you any real benefit.
The link you provide doesn't cover either of these two technologies, so not sure how that comes into play at all.
Just to note it in case anyone comes across it and is confused by the terminology, Cisco did also produce a product called a GigaStack. This GBIC solution really has nothing to do with stacking or clustering, but really only provides interconnectivity between switches.
