This also shows why it might still be useful to include the
original RAW file inside the DNG file. There might be some sort of
information that can be extracted from native RAW files, which
aren't supported by DNG. This could mean loss of data. DNG tries to
get around this, by letting manufacturers add custom meta data. But
not all software that can read DNG can also process this custom
meta data. This might be the problem you are referring to.
Not at all. It's a lot more than a bit of metadata. It's a question of fundamental algorithms.
For example, any raw decoder can handle a basic Bayer pattern raw file with fairly colormetric (responding to color like the human eye) filters and a strong anti-aliasing filter. That got you an acceptable image from the basic Nikon D100 or Canon 10D style camera.
But if you have a weaker AA filter, you now need some sort algorithm to reduce moire (red and blue color bands across detailed areas). If you want to decode a Leica DMR or M8 file, you need an even more aggressive and sophisticated moire reduction algorithm. The DNG file can't provide that algorithm, all it can provide is the metadata that says you need it. So, many raw decoders that can decode DNG from cameras that don't need this algorithm can't decode M8 raw files, even though the M8 outputs DNG as its native format.
There are more severe examples. Fuji S3 and S5 have HDR (high dynamic range) sensors: double Bayer matrix that interleaves red, green, and blue sensors of two different sensitivities. The raw processor has to figure out how to do HDR processing, to put the two differently exposed images back together into one image. DNG can tell you that the data is there, but until the raw processing program knows what to do with the data, DNG is no help.
The Fuji Super CCD sensors (used in the S1, S2, S3, and S5 DSLRs,as well as all their point and shoot and prosumer cameras that have raw output) also require another algorithm to rotate the Bayer matrix 45 degrees, because Fuji rotates it to get higher horizontal and vertical resolution scores.
The best raw processors do sharpening based on the shape that a camera's AA filter "spreads" a point of light into (this is called the PSF, the "point spread function"). To get the detail that Fuji fans have been talking about ever since the S1, you have to cope with Fuji's AA filter having a diamond shaped PSF (again 45 degree rotation), instead of the square PSF of Nikon cameras, or the round PSF of Canons). DNG only codes the "strength" of the AA filter, it says nothing about the PSF shape.
Foveon sensors (Sigma SD9, SD10, SD14, and DP1, Polaroid X530) need unique noise reduction (high correlated noise between channels), clipping compensation, red channel sharpening, and very unique color math because a conventional 3x3 matrix transform (more on those, later) of Foveon data produces strange results (if you want to get technical, the matrix has large off-diagonal coefficients that increase mathematical noise).
Nikon D2X and D200 need algorithms to cope with Nikon's use of independent hardware gain controls on the red, green, and blue channels.
Camera after camera, it's not just data, but special algorithms required to deal with it.
Then there's the issue of color.
All raw files contain very crude color information, just 11 numbers: a red and blue gain value used for white balance, and a 3x3 matrix that transforms sensor output to as close to "human eye" colors as is possible for a simple bit of math like that. This is why 3x3 matrix color tends to leave the reds, blues, and purples a bit unnatural looking. Fuji, Foveon, and Sony's RGBE sensor (only used in the Sony 828, to date) are especially ill suited for standard 3x3 matrix color.
Most raw processing software developers want better color then you can get from a 3x3 matrix operation, so each implements a different way of looking at color. Phase One Capture One is perhaps the most sophisticated, they use two space-to-space interpolators, basically ICC profiles of the sensor in warm and cool light, and interpolate between the interpolators to get good color for whatever lighting conditions you shot in. So, P1C1 can't do raw decoding unless Phase One has actually had a sample of the particular camera in their lab to measure the spectral sensitivity of the sensor and compute those profiles. Converting to DNG won't help. If they've never calibrated the software for a Nikon D3X or Canon 1Ds III, they can't decode the raw files properly.
What sort of color information does DNG contain? Did Adobe look at how other raw converters do color, and make DNG capable of satisfying every raw converter's needs? Did they ask other raw converter writers (you know, like you do if you're really creating an open standard, instead of just making a lot of noise about doing it)? No. They simply gave DNG the kind of spectral coefficients used by Adobe Camera Raw.
So, DNG isn't good for the consumer, it's good for Adobe. It insures that no other raw decoder can ever outperform ACR. Is it any wonder that so few are buying into it?
--
Normally, a signature this small can't open its own jumpgate.
Ciao! Joe
http://www.swissarmyfork.com 
