Re: CFD: zXIF 2017-0207
Glenn Randers-Pehrson <[email protected]> Wed, 8 Feb 2017 19:51:35 -0600
| Newsgroups | gmane.comp.graphics.png.general |
|---|---|
| Message-ID | <CA+PdXctRCu8Bgb8Z3S-OLK8pam9+9SOHue-Fy7rE501UP_NnSg@mail.gmail.com> |
I think I meant "shall not use" when I said "shall ignore" but I agree that it needs to be precisely defined. The main problem with the Exif profile that I see is that it contains several types of tags: 1. tags containing information about the camera. This part is historical data that an editor shouldn't change, and is therefore copy-safe. 2. tags containing information about the picture-taking conditions, also historical copy-safe data. includes GPS info. 3.tags containing information about the scene (subject location, etc) that must be updated when the image is cropped, rotated, or resized. The updating requirements are explained in the Exif spec, Annex E3. This is copy-unsafe, but an application would be allowed to update it by complying with the requirements of E3. 4. tags containing information about the image file (strip layout, compression method, etc.). These don't have any meaning when the image is converted to a new format, but might be of historical value (e.g., was the original image sub-sampled or not?). These should either be copy-unsafe or we could mandate that they be ignored by decoders and discarded by encoders. 5. tags that replicate existing PNG chunks. This is troublesome, when a PNG decoder finds the same data (or conflicting data) in PNG chunks and in the Exif profile. Includes image dimensions, resolution, gamma, sRGB, etc. This is copy-unsafe, ignorable, but we have to allow it in the chunk for the sake of simple encoders that simply dump an existing Exif profile into an eXIF chunk. We should not allow these to serve as falllbacks when the corresponding PNG chunks are missing. In a PNG, no gAMA chunk means gamma is undefined, and does not mean fall back on the Exif data; if we were to allow that, unspecified behavior is unavoidable, as in the scenario Pavel described earlier. Ideally, type 1 and 2 information would be written into a copy-safe PNG chunk (eXIf or zXIf), type 3 stuff would go into a copy-unsafe PNG chunk (eXIF or zXIF). Type 4 information should probably be discarded; even if the image gets converted back to JPEG, the original JPEG information would be obsolete and the encoder should write new JPEG format-related tags. I don't think we want to define two chunks to contain the Exif profile, so it will have to be copy-unsafe. . On Wed, Feb 8, 2017 at 6:20 PM, John Bowler < [email protected]> wrote: > On Tue, Feb 7, 2017 at 8:56 PM, Glenn Randers-Pehrson <[email protected]> > wrote: > > This paragraph, admittedly, is not as clear as I'd like. > That is the challenge; clearly stating exactly what tags are permitted > and when they are permitted. > ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ png-mng-misc mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/png-mng-misc