Home Products Download Order Contacts


Subject: Re: OpenRAW article: "DNG is not the answer"!


The reason I posted what I did is that there is an infinity of possible false assumptions, false incompatibilities, and false dichotomies. We could spend forever if some people listed them, and others responded to them. I am coming to realise how much time I've wasted over the last year or more responding at that level. (Not wasted my time with you, I hasten to add!)

The way to handle these discussions is to develop higher-level models that can be discussed, so that answers in those higher-level discussions respond to an infinity of lower-level queries at a time. These higher-level models go by names such as "computer architectures", etc. I used to be pretty good with computer architectures, (which is why I was appointed a Fellow of the British Computer Society), so I need to get back to my strengths.

I mainly think in diagrams, (and mathematics), so I probably need to devise a suitable set of diagrams. (I doubt if I will get far explaining things in terms of mathematics! Besides, I have never studied the mathematics of digital imaging - I graduated in mathematical physics in 1968, which preceded most of the important work). I wrote a laser printer driver for WordStar 3 (sic) in the early 1980s specifically so that I could draw computer architecture diagrams. Now things are much easier. (Horror - I normally draw with PowerPoint, not Photoshop!)

Here is a TRIVIAL example of the sort of diagram I use. (I used this at my local photographic society). It isn't an answer to any questions here, but simply an example of the way I explain things in diagrammatic form:

Did I succeed? Here is the report:

I probably need to develop a hierarchy of diagrams. Some will be "plan views" - looking at the total digital imaging system "from the top", showing all the important components, such as cameras, raw converters, photo editors, websites, printers, magazines, etc. Some will be "elevation views" - identifying layers. A physical layer, which we can all probably agree is irrelevant to this discussion. Some sort of file encoding level, (TIFF or whatever), which is also insignificant, although some people think it is important. And an image level, which includes the raw image data and the associated metadata. No doubt people will add layers above this, and these may include generic algorithms for handling features at the lower levels, such as special demosaicing algorithms, or XMP namespaces and decoding features.

I'm still thinking about how to represent the time dimension. For archiving, this is obviously vital. It may need a different type of diagram, which anyone could do. The trick is to integrate it with the other diagrams. This is where "self-containment" becomes important.

"the reason some authors haven't supported the Leica DMR is precisely
because of the anti-aliasing issue."

I don't believe that for a second! Frankly, it is utter BS. All they have to do is demosaic the raw image and leave any problems to post-processing. If they say that is what they are doing, what is the issue? The REAL reason that people don't support particular features and particular cameras is "return on investment". Spending one man-month improving support for recent Canon cameras will gain them more customers than spending one man-month supporting minority cameras, so that is what they do.

"that means either DNG is totally self contained, or partially self contained.
since by your own admission you felt the underlying system technologies
would need to be modified for new "sensor configurations", i think it's
extremely reasonable, and probable, to conclude that DNG is not totally
self contained."

DNG is self-contained. You don't need to know details about the camera model, or details about specific cameras within that model, to process the DNG file and reveal the image. The fact that "underlying system technologies would need to be modified for new "sensor configurations"" is TOTALLY irrelevant to this discussion!

For example, the current DNG specification ( caters for Bayer sensors, without needing to know details about the camera model, or details about specific cameras within that model. (Anyone who disputes this needs to ask themselves how ACR 2.4 can support all those new cameras without having to be updated).

It may need a new DNG version (perhaps to support sensors arranged in concentric circles. Let's suppose such cameras appear, and a new version of DNG is published to cater for that characteristic. Version will then be self-contained for all cameras that use Bayer sensors or the new sensors. Version will STILL be self-contained for all cameras that use Bayer sensors. BOTH versions will be self-contained - but within different scopes. ALWAYS SELF-CONTAINED!

"this suggests the burden of interpretation is placed onto the underlying
engine. if that's the case, it's legitimate for developers to question
what distinguishes DNG from other containers? Yes it's a well organized
and predictable container, but one that still contains fields which may
or may not make sense to the underlying engine."

DNG is a raw file format. (It is a specification for a passive, one-direction, non-interactive, interface). Why shouldn't it place "the burden of interpretation ... onto the underlying engine"? DNG isn't magic! It still needs raw converters to do significant things, such as "interpretation"! Repeat: "DNG is a raw file format"! It is a way of carrying raw image data from the source of that data to a place where it can be converted into the next stage, such as a rectilinear RGB representation of the original scene.

THE significance of raw is that it delays this interpretation, places the responsibility for it onto a more powerful computer, and enables it to be redone at will:

The significance of DNG is that it is by far the best raw file format on the planet. But: "DNG is a raw file format"!

"which leads me to question why, when advocating DNG in many of the online
forums, here and elsewhere, you've proposed a major benefit of DNG is
that it's immediately compatible with existing solutions--namely ACR."

I don't do that! Where DNG is concerned, I consider that ACR is just one of many raw converters, now and in the future. It happens to be a good example of how DNG can be handled, but Silkypix is another interesting example. ACR 2.4 is an EXAMPLE of the power of DNG - it isn't a reason for DNG.

"it's important to understand this may be why there has been so much (perhaps
legitimate?) resistance to the unfettered adoption of DNG."

It takes effort to support DNG. It takes even more effort to support it properly. Products suppliers need excuses for their customers. Different product suppliers provide different excuses. Raw Magick say it would inhibit innovation. Bibble say they need an SDK that would enable them to provide a DNG converter (why?). Pixmantec say they have better things to spend their effort on - honest! Phase One say it is a business management decision - also honest! I expect Pentax to be sensible. I hope that Canon will be too. I expect Nikon to be stupid. All of these are glimpses of the truth - it is a business decision based on an evaluation of return on investment.

"what makes sense to me, is the pursuit of a total, or near total self-contained
format. one that might function in a similar fashion to the typical color
management model. the file format would contain a specific, but portable
appearance that when sent to any raw processor would attempt to preserve
that appearance."

Digital still photographs don't resemble the real world. I've been in the audience at 3D IMAX presentations, and they are vastly closer to the real world than digital still photographs. In virtually every dimens


View All Messages in adobe.digital.negative

OpenRAW article: "DNG is not the answer"! =>


Copyright 2006 WatermarkFactory.com. All Rights Reserved.