So here we go again. I don't know how many times we have to have this discussion, but it seems that we keep having it. Which is better, compressed or non-compressed? Bit reduction or full bit recording?
Back in the early days of DSLRs I used to fight with some other well known big name photographers who went around saying "you only need to shoot JPEG." Every time they said that, I would have to go around saying "raw is better." Funny thing is, every one of those early JPEG pundits is now using raw. Why? The explanation is simple: data reduction can't be retreived. You're always better off with more and more accurate sampling data than data that's been through the wringer.
The "evidence" for "JPEG is good enough" is always the same: "I can't see any difference." Well, that's the point. JPEG (and MPEG which I'll get to) were about data reduction that didn't become visible to the majority of people. The data reduction is still there, though, and it'll nip you by your short hairs the minute you start trying to do much in the way of post processing on the data. 8 bits is not 14 bits. sRGB is not ProPhotoRGB. Fourier reduction of pixels to formula is data reduction. Lost data, lost data, lost data. You never get this data back, and some day you'll find that you want to make a change that requires it. Sure, most of us have learned how to patch over JPEG issues in post, but it's more work and there will always be things you can't do because the data just isn't there.
Don't get me wrong, JPEG is indeed "good enough" for most people, and a lot of pro work, too. But if there's a chance that anyone is ever going to touch your files after capture and make changes to pixel data, not having missing data is really, really helpful. So helpful, that the first time you encounter what missing data does to you in post processing, you'll start using raw. Heck, you still get a JPEG every time you capture raw, so if that's what you want, have at the embedded image.
Lately we're hearing the same argument all over again: "you only need to shoot video in AVCHD." The well known pros saying this are using the same old "good enough" and "can't see a difference" arguments. Apparently they've never had anyone seriously color grade (post process) their video to any extreme. Because missing data is still missing data. Again, we get bit reduction (typically 10 bits is being dropped to 8 bits). We get color reduction (4:4:2 or 4:2:2 becomes 4:2:0). We get completely missing frames (we're doing a JPEG-like compression only on the data that changes between frames), and we get variable compression levels (bit rates).
Video is no different than still photography. Indeed, I try to approach both similarly: collect optimal data. Make the data neutral and flat. Don't put compression or other data reductions in place until you've made all your pixel level corrections (which includes post processing). No difference. Capture flat, neutral raws with no highlight blowouts in stills; capture flat, neutral uncompressed 4:2:2 or better (preferably in 10-bit) with no highlight blowouts in video. Post process the proper contrast and other effects into both. Post processing isn't a chore, because all the data is there and you're not dealing with gaps. Simple as that. Yet we still have pros saying "AVCHD is good enough." I'm betting that in a few years they're all recording with less or no compression.
So let me predict the future. Some day we'll have holographic recording at the consumer level. Because there's a lot of data involved, the holomakers will come up with HPEG. Early pro pundits will go around touting "you only need to record in HPEG" because they "can't see a difference." Meanwhile, the rest of us who know better will be recording in HoloRaw. And when people start asking why our work looks so good and how we managed to pull out highlight and shadow and 3D detail, guess what my answer will be? ;~) Heck, guess what Hollywood will be doing?
Record optimal data. Keep that data intact for as long in the processing as possible. Reduce data only when the output format is known, and optimize for that format at output only.
Of course, the masses will continue to use JPEG/MPEG/HPEG and be perfectly happy. But you're not the masses, are you? You're the 1% of photographers. You know what to do. (No, the answer isn't Occupy Sony.)
Okay, maybe some of you still don't know the right answer, so let me make it personal. I just invented a cloning machine. It can only make its copy of you just before you die (exactly when you most want to be cloned, memories and all). I've put two recording settings on my machine: 8-bit DNAPEG compression and 32-bit CelluRaw. Which setting would you like me to use for your clone?
What makes me think some pundit is going to say "DNAPEG is all you need?" Well, at least you'll save on memory (see next story).