When I came back to photography, the film workflow I cut my teeth on was supplanted by digital, so I had to figure that out. I started just shooting to JPEG, rapidly found myself trying to mung the noise out of and sharpness into my captures from a 6MP camera. Generally got good captures right out of the camera, but when things went south of that…
The picayune thing that finally pushed me to raw captures was the desire to make “proof-sized” JPEGs right off the bat. For me, proof-size was (and still is) 800x600, and no Nikon camera (another constraint to my consideration, started with Nikon in the '70s, inertia…) that made such small renditions. That, and my aerospace-influenced way of looking at data, that the raw telemetry is sacrosanct, gravitated my sorry behind to the concept that the raw file is the fount of all rendition.
That thinking greatly influenced my development of rawproc, very much oriented to the efficient processing of raw originals. Later, with a better camera and the blessing that is highlight-weighted metering, that workflow enabled dealing wholesale with the captures from a scene; from the proofs, re-process the first capture from the scene, delete the other proofs, and re-run batch with the first-capture’s toolchain. Yay.
This thinking also influenced my perspective on metadata. That is, the metadata in the raw file is associated with 1) describing the raw file, and 2) what’s needed to process the raw file to rendition. With the glaring exception of the camera color matrix Nikon supports this, so the raw file’s metadata is where I go to look for that stuff. I particularly don’t wholesale copy all that to any rendition, too much futzing for stuff I can get from the original raw. All I send to renditions are the essential capture information: f-stop, shutter speed, iso, focal length. Easily handled with the libjpeg/png/tiff libraries, so I recently removed exiv2 from the essential rawproc build. However, I’m starting to cogitate how to use the Nikon Z-mount lens correction data, and that bolloxes the above. Might try to talk libraw folk into including it in their data presentation…
Anyhoo, coffee-fueled Sunday cogitation …