A virtual DOM is essentially a data structure that abstracts over modifications to another data structure simply for performance reasons. It's just a way of modifying some data in memory and then committing all of those changes in a single transaction. The very fact that this is actually a reasonable solution given the tools available just goes to show how broken the tools are. The whole computational model exposed by browsers is completely out of touch with how it's being used. For example, why isn't there a language construct in JavaScript inside of which all DOM manipulation isn't propagated to the rendering engine? Instead we build a virtual system inside of a virtual machine to achieve the same, and then wonder why things are so clunky.