[Radim]

The lessons you learn on your own stick the best, don't they?

I remember a similar "heureka" moment when I derived a line-drawing algorithm that respected "subpixel" starting and ending points. That is, don't assume a line starts/ends in the middle of a pixel (like most algos do), but rather at an arbitrary point.

I forgot why I needed this (something to do with near-axis-aligned polygon slopes not looking right on a highly rasterized problem). But I do remember the elation when I finally got it to work :-) Also early 1990s, no internet and no English for extra fun.