Technically it's three trials of 2D data, with the axes being Time and Memory Usage. The width of the lines given by the version axis is for visibility, not to encode data.
A 2D line graph with 3 colors for the different versions would be an appropriate visualization. This one suffers spatial distortion and is only useful in confirming that yes, they use less memory in step 2 now.
>The width of the lines given by the version axis is for visibility, not to encode data.
Bar graphs do exactly the same thing in a 2D graph. Would you suggest that all bar graphs be replaced by a one pixel wide stacked bar graph?
Of course not, because even though the samples in a bar graph are discrete, they convey order. It's the same here. The width of the lines is only for visibility, but the Z position of the lines conveys version order. A 2D line graph would only convey that by labels. And if labels were as useful as visual position, we could just replace all graphs with lists of numbers.
I personally found this graph trivial to read and comprehend. The spacial distortion is canceled out by the fact that the differences we're looking at are very large, and the steps axis is spaced out sufficiently to line up the points visually.
My only complaint is that it shouldn't be a line graph, as the steps are not continuous.
I do agree that a style with discrete steps would also be appropriate.
To the other points, no, not all bar graphs do the same thing. Bar graph bars are often the width they are due to data binning, where the width has significance.
The graph was trivial, but made less clear by the 3D isometric projection. the spatial distortion is reduced but not canceled out by the space.
3D pie charts share the same failure. Yes, as long as the tilt is slight, they are still mostly readable. But it will make "closer" wedges appear larger and "farther" wedges appear smaller. It's visual dishonesty and unnecessary.
Just because one can finagle understanding out of a visualization does not make it a good visualization.
A histogram of a normal distribution (say, mean lifetime of a particle) is a 1D histogram. You would have counts of decays vs time. In your example, that would be 2D, but it's not.
A 2D histogram would be more like a heat map. The X axis would be time, y axis version, and the (x, y) value would be memory.
But I agree there may be a better way of visualising this data.