Al Dykes wrote:
[snip]
Aperature looks like, first, a workflow tool and the comparison to Bridge (whick I only know from Apple demos) and C1 is valid.
I’ve been thinking about what Apple could achieve using their approach of non-destructive adjustments to raw files. I’ve come to the conclusion that they could actually develop Aperture into a comprehensive photo-editing tool that would be very serious competition to ALL of Bridge + ACR + Photoshop. In much less time than it took to develop Photoshop to its current state.
Here are two "paradigms" for holding photo-edits here. One is to hold "instructions", the other is to hold "results".
Photoshop started life holding "results". It held a digital image as a rectilinear array of pixels. As you edited, it changed those pixels. You saved the results of the editing as changed pixels. Over the years, it added the ability also to hold "instructions". An adjustment layer is a set of instructions for changing an image. It doesn’t actually change the pixels directly. (They get changed temporarily when you print or display on a screen, or permanently when you flatten the layers). So Photoshop has become a hybrid, but sometimes a bit uneasy.
ACR, Aperture, and I believe LightZone, don’t save the results. They save the instructions. This is potentially very powerful. The downside is that it needs a lot of processing power, because, to use the results of lots of editing, it has to apply all of the instructions so far, instead of the just loading up the changed pixels from previous editing. That processing power is becoming available, and is what Aperture exploits.
Examples of the differences between these two paradigms:
To crop in Photoshop, it basically discards the unwanted pixels, and the result is the rest of the pixels. To crop in ACR, it creates simple XMP metadata describing the coordinates of the corners of the crop, and leaves the image data unchanged.
To do red-eye in Photoshop, it changes the values of the pixels concerned in the image array. If ACR did red-eye, (which Nikon Capture and Aperture do, so I expect ACR will in future), it just needs to hold about 4 numbers: the x & y coordinates, and the pupil size and darkening amount, in XMP.
In principle, just about ANY editing should be possible using the "instructions" paradigm. After all, you could record everything you did with Photoshop, then instead of saving the "results" (as pixels), save the record of the instructions. These could be applied later to the original image. So if the instruction-metadata also holds the order of editing operations, it can in principle to do anything Photoshop can do. But it would use one single model, not Photoshop’s hybrid approach. Potentially, much simpler.
Once it is the instructions that are saved, there are some super possibilities. Think about the way that lots of HTML web pages can all use a single CSS file as their style sheet. Change that CSS, and all of those pages change their rendition without changing their data. Imagine lots of image all refering to a single instance (not multiple copies) of metadata. Edit one of those images, that metadata changes, and all of the images change simultaneously. This is the sort of thing that is becoming necessary when handling 100s of raw images at a time. With ACR, and probably with Aperture at the moment, you have to apply the change to every image, perhaps by some sort of copying operation. But, in future, by sharing metadata, changing one bit of data would change them all. All the computer science concepts of OO, type hierarchies, and inheritance, could be used to make the workflow easier.
It is clear from Apple forums that Aperture is just at version 1, and they are already working on later versions. I hope Adobe are working on the enhancements to Bridge+ACR to achieve similar capability.
—
Barry Pearson
http://www.barry.pearson.name/photography/ http://www.birdsandanimals.info/ 
