Originally posted by newmikey I really shouldn't be saying this as I simply do not know LR at all but, isn't it that regardless of the output format, the content of any such file can no longer be raw sensor data, or do I simply misinterpret the use of the word "raw" here to indicate something totally different?
I had the same issue with "layering raw", raw data, pre-debayering, has no dimensions or pixels which can interact with pies on other layers. Maybe the software creates a pixel matrix (bmp or tiff) in memory on-the-fly, but it can certainly not be acting on a raw file itself.
Last question maybe for the pundits: raw is raw because it is a one-way format: you can copy the raw sensor data, extract the image information and pixels from it but you can never create a true raw file without clicking the shutter. I suppose I don't really understand what "saving as DNG" really means to the raw content of that file. (I never did accept the "DNG as savior" and stuck to PEF throughout the years). It seems even Pentax is still strongly hanging on to PEF, even with the K-1.
A RAW file is a lot of stuff. Some image stuff; some metadata stuff (like time, GPS, etc). It's a lot of values for various different parameters; Olympus even includes corrections for its lenses. And there's a JPEG preview too, the thing you see in the camera when you review a picture. A Pentax pixel shift image includes four different shots. DNG takes that a big further, and includes the metadata that is normally stored in sidecars, like IPTC and so on.
Lr is a parametric editor. ANY time you edit, it stores that info in its database. So if you apply X amount of sharpening to a PS DNG, it stores those instructions in its database. The RAW DNG is still the same. But you can also write that parameter, X amount of sharpening, to the DNG file, and other applications, like Mylio, Photoshop, Bridge, etc can see that and apply the same parameter.
To make a panorama or HDR merge, however, Lr or Ps has to go further, and actually create a new file (ditto for DxO, Macphun and maybe others). That's because it is actually blending information together, and rather than just registering light in a mosaic, it's representing it more dynamically, as explained above. See here for more info:
http://protogtech.com/adobe-lightroom/adobe-dng-hdr-format-part-2/
But then that info is left in a new DNG; it's a different type of data, but still includes all sorts of that other stuff the different DNGs used to make it had, like lens info. So you can apply lens corrections. And all sorts of other adjustments, just as you would with a sensor-created RAW. The Adobe engineer joked that since they are closer to RAW, having not been developed, but not mosaic data, they should be called "rares." Pretty apt.