Correct...if you open the image into Photoshop, it's rendered as an RGB file. If you open as an object, the raw smart object is yet to actually be rendered until you either flatten of collapse the smart object. Note, the raw SO embeds a copy of the original raw file inside of the PSD or TIFF in Photoshop. So, the resulting Photoshop document can get quite large. You also sever the .xmp settings from the original raw file so if you make changes to the embedded raw file, it's not easy to get those changes updated into the original raw file.
Also note that there are some things you can't do with a Smart Object such as paint or clone on the SO. You can make a copy and flatten the SO so the copy behaves like a normal pixel layer.
Raw Smart Objects can indeed be pretty handy...perticularly if you want to blend between 2 or more sets of image ajustments but it does come at the price of file size...so i would only use it when you get the full benefits and not when simply doing raw processing.
Thanks, Jeff. That's just the answer I was looking for!
One question remains: As you explained, when I use Open Image in ACR, a rendered image is sent to PS. But if I subsequently do Filter -> Camera Raw Filter... in Photoshop, it launches ACR. Without the raw data, how does ACR work? Does it simulate working with raw data? And if so, are the adjustments less powerful then they would be if operating on raw data?
If you are talking about ACR as an adjustment filter in Photoshop CC, yes, it's working on the rendered image not the raw image. So, it will have some limitations-particularly with highlight recovery.
Exactly! Thanks again.