My understanding is that this paper provides a new (albeit as early as 1993) solution to the selection problem.
Before this paper came out, for example, to change the color of a certain object, we have to click and select that object, or drag and highlight a line of text, or drag and group a set of objects, followed by picking color, confirming, etc. There are at least two problems in this conventional selection method. Firstly, it is not easy to select. The size, shape, color, etc. of objects add to the difficulty of selecting them. For example, even with the latest version of Photoshop, it is still quite tricky to select, say, the face of a person. Secondly, it is not convenient to do the following steps. Spatially usually we need to travel to somewhere like a panel to change the color, font styles, etc. Microsoft Office offers a close-up widget available instantly after you select some text. This idea is similar to Toolglass and Magic Lenses. Reducing the spatial distance can reduce multiple steps.
Hence, to solve the selection problem, this paper offers a two-step approach, namely, Toolglass and Magic Lenses. Toolglass is like a transparent layer between application objects and conventional user cursor. This layer can help us make selection by, for instance, amplifying local area or reducing visual cluster. The second step is called Magic Lenses, which makes the following steps readily at hand (close to the layer). Compared to drag and group and then clicking on a side panel, this two-step approach is more efficient and user-friendly.
What is similar to this framework is the focus+context concept in visualization. I would say Toolglass and Magic Lenses express a more general concept including some of the ideas in visualization, e.g. amplifying local area. However, what it fails to address is offering context. Judged from the paper, the Toolglass and Magic Lenses result in a discrete two-layered view of the application – users might lose the context.