Media Settings:
You are correct in that the epson driver is black box. It comes preloaded with the media settings for espon media, and some generic media settings for matte and gloss papers that already have the ink limits, overlaps, and linearization premade, which is why ICC profiles are so important with color printing to correct for these generic settings. When creating QTR profiles what you are actually doing is creating custom media settings for each paper that is used. The better the media settings, the better the print (without color management), and the less need to use an ICC profile to correct for errors in the gray curve or final linearization. The screen to print match is a different issue entirely.
Print Drivers:
QTR isn't analogous to a print driver, it is a print driver, on both Windows and Mac. Mac uses a standardized printing system (CUPS) that allows it to be used across the whole system installed printers on the system. Windows uses something else for printing (I'm not sure what that is), so an entire application needed to be developed that contained the print driver and spooler (the guten-print core that was used for QTR), media settings generator (Roy's curve creation programs), and a place to hold the file that is being printed. I'm not a developer so I could be wrong on some of those details about how and why certain things ended up as they are on the different platforms. The point is I see the QTRgui is a "compromise application" and can be very limiting compared to printing on a Mac.
Curves and Descriptor Files:
The ink descriptor file can be roughly thought of as "describing" the curves. However saying that they just contain basic parameters can be a little misleading. They are detailed sets instructions for how the curve generation program creates the overlapping ink curves in the .quad file. That might sound like nit picking, but I think it is an important distinction in that the ink descriptor files can be very precise in controlling how/when the inks ramp up, the shape each curve takes, and the final linearization. What you wrote was a good basic explanation of what is happening, but sometimes those little details that are missing from the basic explanations are the things that end up holding people up and getting stuck in the process.
Richard Boutwell