In a medium complex drawing it may be too much details to have all layers in a drawing visible at once.

Instead a number of views would be created that saves not only zoom level and position but also what layers are visible and not (as appropriate for the purpose of that particular view). This is very cumbersome (to the extent it is not really realistic) to achieve today and requires a lot of duplication of both model and layers.

The advantage of adding this is the ability to have only one model (with multiple views) and avoid duplication of work and the risk of inconsistency inherent with many copies of the model.

As a further (and natural extension to this) would be the ability to used this functionality with view-ports to be able to create (possibly many) paper-spaces from a single model. This would however require view-ports to be aware of what view they are a window into. A single change in the model would then propagate to all defined view-ports and paper-spaces.