Hi Bill,

First of all, thanks for the detailed description and samples. It makes diagnosing tough problems like this a lot easier.

I had a close look at the DxO and exiftool samples, and as far as I can see the exiftool image contains all of the information in the DxO image. However, the one difference (and this is a difference that may confuse some EXIF readers) is that the exiftool EXIF information is written in little-endian (Intel) byte order, while the DxO image is written in big-endian (Motorola) order. But both maker notes are written in big-endian order since this is the way they are stored in the NEF. It is possible that an inflexible reader could have problems with big-endian maker notes in a little-endian EXIF image. This really should be considered a reader bug, because this sort of thing does happen occasionally, and there is a separate byte order mark in the EXIF and maker notes so the individual ordering can easily be recognized.

The reason ExifTool writes in little-endian order is because that is how the EXIF data is ordered in the source image dsc_4714_rse.jpg. ExifTool will not change the existing byte order, but if it creates the EXIF from scratch it will default to the same order as the maker notes. So if you delete all of the EXIF and rewrite it completely, you will have everything in big-endian order.

To test this theory, try creating another version of the image with exiftool. First copy "dsc_4714_exiftool.jpg" to "dsc_4714_exiftool_2.jpg", then do this:

If this works, try sending a bug report and the original exiftool image to PTLens and irfanview to see if they can support this type of mixed-order image.

- Phil