He did mention that release of 1.4 is nowhere close to being ready.
I read the included link, and even the link within the link, and stuff does get over my head pretty quickly. Especially his conflation of true drizzle and CFA (or Bayer) drizzle. Though PI may be similar?
My impression has always been that "true" drizzle is essentially an upscaling, but the data rains down onto a finer grid. If you are undersampled, this can produce enhanced resolution when everything is stacked. If you aren't, then there is no recovery of resolution, though I don't know if still provides an "appearance" of being smoother or more rounded off when it comes to things like stars.
True drizzle could certainly be run on OSC data. I though, maybe wrongly, that it was done post-debayering.
Bayer drizzle instead replaces debayering, using the shifts, drift, and dither to fill in the entire grid with actual captured data, in three channels across the entire image dimension. The point being to avoid artifacts inherent in various debayering algorithms, as well as just knowing that everything is a real photon. Absent any error handling that may be needed, of course.
DSS says the concepts are related, but warns that you should not do both. But maybe they are the outlier?
Perhaps it all comes down to implementation, particularly as to star registration because sub-pixel transforms are required (splining, warping?) that have to go beyond simple dX dY and rotation angle. Maybe that's why they would end up conflated or the same?
