This is a very broad, sweeping, unsupported swipe at one of the deepest and most sophisticated imaging applications ever devised by mankind. Could you please substantiate what you are saying? And who is providing equivalent functionality in a manner that is less "obsolete"? Note that I do not work for Adobe, I have no particular relationship with Adobe except as a user of their software, but when I see comments of this ilk it leaves me kind of breathless.
While I might think your assessment was at least as sweeping as mine, I agree your question deserves an answer.
Photoshop uses a one-dimensional dataflow architecture known as "layers" -- perhaps one and a half dimensions when you count things like masks. Compare this with an N-dimensional dataflow architecture such as Nuke. Compare with respect to a common task, such as compositing.
Compositing is inherently an N-dimensional process. Several sources, each processed differently, must be brought together into one image, with integration occurring potentially at different stages in processing for each source.
- In photoshop, it is necessary to "bake in" intermediate results in order to align them to a one-dimensional flow. There is no going back to revise parts of your composited image, all of which might have required extensive independent treatments, as well as a level of /coordination/.
- In Nuke, as with any N-dimensional dataflow (or "node based" editing if you prefer), it is possible to do several things:
(1) You can import and/or retain the entire processing history of the multiple sources and revise at will at late stages in the project.
(2) One source file can be used in multiple stages of the dataflow for entirely different purposes without duplication or baking in.
- Using an N-dimensional dataflow machine, you can devise many different kinds of user interfaces at the level above, and that includes a photoshop compatibility layer, as well as other things for various different purposes.
- In photoshop, several hacks have been devised in order to accommodate different needs, such as "Apply Image..." which is completely unnecessary in an N-dimensional dataflow architecture.
And that's just a start. Thanks for the question though. I'm thinking of teaching a course on this over the summer. It's a timely topic.