with the Subdivide Collapse with Relative Details, i understand the details size in pixels to be viewport based, so thats ok.
what gets a bit confusing is setting it up for Subdivide Collapse and Constant Detail. whereby as I understood when you were walking through it the percentage was based on a Blender unit. so 5% would be a detail size of 5 percent of a single blender unit, and increasing that percentage would increase the edge length to use.
but, ive noticed that in 2.79 its no longer a percentage, its labelled as resolution, which seems to have an inverse behaviour to that of the % of blender unit constant, whereby higher resolution higher poly count.
Just thought id mention it, as it confused the heck out of me until I read the tooltip and had a play around with it ;)
