it took 2 or 3 solid days of working to get it all working.. half a day to most of it, and then another 2.5 days of trying to get it working as well as i could and tweaking. i'm sure we will still find some minor issues but i'm glad it's basically working! thanks to everyone for being patient and reminding me this feature was needed.
note to coders 1: one lesson learned is that Win7/Aero glass on vmware operates in non-accelerated non-3dgraphics-card-simulation basic-theme mode, which CANNOT reproduce the semi-transparency (at least my vmware 6 can't, maybe newer ones can), and so is not useful for testing such stuff -- that's why i thought for the longest time that this corner transparency issue was already solved.
note to coders 2: i give credit in the help file but i should say so here as well, that Ian Griffiths's write-up of capturing transparency provided the eureka moment for me on how to get transparency information (and convinced me that trying to find a set of windows api functions to do it was hopeless):
http://www.interact-...7/01/30/recoveralpha ; a word of wisdom to those coders trying to use this method -- it won't work perfectly as described on that page, because Aero glass effects don't do a pure linear alpha blending.. unless you can figure out the exact blending algorithm MS used (i couldn't) you are going to have to come up with some heuristics for recovering the underlying transparency and source colors (one thing i did was solve for the white-background pixels instead of the black ones since those seemed less muted).