I think, it will make sense to present this from an engineer's point of view. Not because most people in this argument are engineers but because I think, it's engineers' thinking spilling into the mindset of people who are not engineers and can't apply or explain those ideas clearly.
First of all, for an engineer there is nothing unusual with various types of technology replacing each other. The device that some company is producing now, likely has a design based on some decisions made years, possibly decades ago, based on technologies, components and materials available then. Company does its own R&D. So do its suppliers. At some point every part that company uses, be it produced by them or someone else, will be out of production, replaced by something else -- possibly based on a completely different technology. Even if it won't be, engineer himself may find a better way of doing something, and eventually he will want to get rid of something that possibly was a central piece of the product line when it was introduced. So engineer is constantly (and often painfully) aware of the changes that happen around him, changes that may show up in production years later but for him they are happening at the moment he compared datasheets or read a paper, and decided that some serious change is coming. The products may be still coming out of the assembly line, but engineer knows that parts of their design are for all practical purposes dead, never to be encountered again, save for some unusual turn of events when even more advanced technology will for some reason call for an old solution.
Now, please try to understand how such engineer sees DSLRs. Film SLRs are a great, probably the greatest achievement of devices for film photography. Rolls of film can't be easily moved or replaced with a focusing screen -- even if it was possible, nothing can move film at sufficient speed without turning the whole device into a movie camera. For a small fraction of weight, space and energy necessary for placing any other focusing device behind the lens, mirror allows the camera designer to place a separate focusing screen -- with optical and electronic focus aids, exposure meters, etc. in the camera, and project the exact equivalent of the image that will be projected onto the film. The size of the mirror chamber, mechanism, pentaprism, etc. is small compared to any other thing that would allow to reproduce measurements and view with comparable precision.
There may be other solutions and workarounds -- rangefinders with their less precise and indirect measurements, DLRs with secondary lens, large format cameras where film is actually replaced with the focusing screen, etc., but nothing compares to the elegance of film SLR. If it's not perfect, it's very close.
Now, digital cameras happened.
For them, the problem that SLR was solving, never existed in the first place. Remember what it was -- the impossibility of quickly switching between film and focusing screen in the same position behind the lens. Digital sensors always were usable as the focusing screen. The details could be unclear at first -- if CDAF is enough (or what amount of CPU performance will make it enough), what it would take to add PDAF, how to present a high-resolution image to the photographer with limited space of LCD screen. But one thing was certain -- once sensors are not film, and can be engineered with all kinds of weird things on them, the solution will be entirely there. If someone cared, it's possible even to make the sensor translucent and use it as the actual focusing screen -- but likely no one will actually have to go that far.
So never in an engineer's mind it would ever appear that DSLR is a viable long-term solution. The idea is stupid -- no, beyond stupid, it ignores fundamental nature of the problem and physical properties of the materials available for the digital cameras.
In the short term, of course, things were completely different. First digital sensors were awful. Low resolution inherited from TV. Noise. Processing power was pretty bad as well. To focus those things with any precision based on CCD's own output, would be madness. Asking for additional subpixel things when a crude grid of CCD barely can drag charges toward ADC, would undermine the efforts of people who were still trying to get decent resolution out of them. So the old solution, SLR, was the best option available. After all, mass production of moving mirrors, was a long solved problem for the camera companies. DSLRs (and before them, digital backs for film SLRs) became flagships of the new field of professional digital cameras.
The idea of sensor being potentially a better focusing screen than the focusing screen, was still in the minds of engineers, it's just sensors were too bad for this task. But then sensors started getting better and better. Clumsy CCD was replaced with basically IC-on-a-sensor CMOS. Measurements became more precise, and adding weird things to the grid of pixel was no longer an insane thing to ask for. Processors, thank to cell phones, became more powerful as well. Not only large "professional" cameras were getting those benefits, but when users of P&S cameras and phones were satisfied with enormous numbers of poorly exposed pixels with infinite depth of field, same technology was ready to fulfill the promise of "better focusing screen than focusing screen".
And here you have the modern MILC. CDAF exploits humongous processing power of the camera's CPU/DSP. On-sensor PDAF became possible with sophisticated high-resolution sensors with additional pixels. While this technology is still young, processors may be still slow, sensors may be still more of a grid less of a focusing device, the direction that was clear from the very start, is showing itself -- the last excuses for the mirror mechanism are disappearing, and on-sensor focusing will soon overtake the DSLR autofocus. Oh, and with manual focus things are even better -- EVF resolution is growing, and eventually will reach the resolution of the sensor itself. Better focusing screen than focusing screen, for humans (brighter image, focus peaking indicators) and machines (in-sensor PDAF).
Now, what will an engineer think of DSLRs, other than they are dead -- if not "dead" as lifeless chunks of metal, plastic and glass, then at least "dead end" in the evolution of the camera technology? Of course, the same engineer will know that just because technology is dead, it does not mean that products that use it, are useless or inferior to ones that use potentially better technology. DSLRs still have faster autofocus, so wherever that is necessary, they are the best tool for the job. On top of that, companies with very long history of making high-quality equipment, are interested in prolonging the life of DSLR because those things are cheaper for them to develop to the specs that they achieve at their top of the line products. And as many examples show, few large companies are often capable of freezing and ever reversing the progress in technology for years.
But the direction of the progress can't be easily changed -- DSLR was always a losing combination of technologies that existed because sensors were bad and processors were slow, and no one likes bad sensors and slow processors, so at some point everyone, even Canon and Nikon, will have to face the question -- why the Hell do we still have that mirror in there??! And then there will be no DSLRs left.
So engineer may go to the store and buy D800, fully aware that in a few years it will be not just obsolete as a product, but obsolete as the whole direction of technology. He will want D800 because he has a kid and a dog who like running in unpredictable patterns, and being an engineer and not a photographer he will find it easier to rely on fast autofocus than to go through the trouble of prefocusing, planning depth of field, etc. At the store he will find a professional fashion photographer buying the same D800 because it's Nikon and because it's DSLR. And will think, professional photographer has worse reason for that choice than he does. And the engineer would be right about that.
He will also see Richard there, buying exactly the same camera to take pictures at the sporting events. The engineer will think, Richard has a better reason for buying DSLR than he does, but he will also know that Richard is a troll because of this thread:
http://www.dpreview.com/forums/thread/3593986 , so Richard will never know engineer's opinion about his choice because engineer knows better than to feed the trolls.
But then, engineer will find a bunch of threads where people are proclaiming greatness and "professional" status of DSLRs, and will post his opinion about perspectives of that technology, and areas where it is already being overtaken by mirrorless cameras. Since no one understands details of what engineers post, he will be called a mirrorless fanboi. By Richard.
