"semantic zoom"

[gordonwoodhull]

This term is somewhat overloaded but I'm using it in the way it is sometimes applied to maps, where text and markers stay the same size but the geography grows and shrinks. (Roads also scale at a different rate from the zoom.)

In other words, the ability to zoom the layout while keeping the nodes the same size, providing more space or less space for edges.

It would be too slow to tell the layout about this and we also don't want the layout to change in reaction. So it means the layout would be sort of disconnected from the scale.

Conceivably there could be two (or more?) stages:

1. zoomed "out" far enough that node sizes matter. layout algorithm has to get involved
2. zoomed "in" far enough that layout algorithm can consider nodes as points

I have only started to think this through. The idea seems fairly intuitive and the implementation of stage 2 would be pretty easy. It's stage 1 that causes a lot of trouble (and too much lag for interactive zooming).