Update: This post has attracted a lot of interest, and I'd like to just link to a few resources that are very interesting for "what-should-I-buy?" issues: Snapsort's comparison page is breezy and brilliant for quick camera rundowns. (example: Pentax K-x vs. Sony NEX-3). DP Review is one of the best all-around camera review sites, and while I once teased them about their skewed rating system, they've revised it, making their work even more useful. I also owe Tim Bray a link, because while my allegiance to Pentax gear started with my father's Spotmatic (he had two bodies, each with his name and address on a piece of paper taped to the bottom plate; it was a proud moment when he put my name on the bottom of one of them), Tim is a visible, local proponent of the Pentax pancakes, my favorite lenses that I don't yet own.
In the long run (ten years?) the cameras available for sale, at both the pro and consumer levels, will be virtually unrecognizable. I'm not even up to date on the latest in imaging research, but expect major weirdness, driven by applying a LOT of computing power to light-gathering systems that act nothing like current lens-and-sensor designs.
In the short run, a few predictions that are semi-useful camera buying advice. Take this with the grain of salt that I just bought a DSLR with a movie mode, and think it's very nice. Take it also with the disclaimer that I'm not a pro photographer.
The Disappearance of Compact Cameras
The most obvious of my predictions is this: current mobile phone cameras are so good that they are obsoleting the idea of a separate very small camera. The dedicated small cameras are still better cameras (I use and like a Canon SD1000), but they don't offer sufficient benefits over mobile phone cameras to carry a second device for candid shooting. The last small cameras that will survive in any volume will be a merger of video camcorders and superzoom compacts, where the long and versatile lenses will continue to sell to parents taking video of soccer games and piano recitals.
The corollary is that mobile phone cameras will get better, but within the size limitations of the phone. A lot of clever hardware and software will be thrown at mobile phone cameras, no surprise there.
Oh, and as a small sideline, the market for mini video cameras like the Flip is obviously dead, except for sporty nerds like me who need cheap video cameras we can put in harm's way.
The Pro Market
Mostly stasis. Full-frame cameras (and the poster child is the Canon 5D Mark II) are the workhorses here for pro shooters, and the split between Canon and Nikon is relatively even and most of the market (Sony makes full-frame DSLRs, but they're not yet a big player. The offerings are decent; maybe they will capture some market share). Nice APS-C cameras are used by a certain number of pros (think Canon 7D and Nikon D300, but also a few weirdos with Pentax K-5s or Sonys), and they work. Some pros are even shooting with micro-4/3s cameras when size really matters, as if the pro rangefinder market had woken up after 50 years of sleep.
I don't see this changing. The 5D Mark II (and the similar Nikon D700) will probably be the Standard Default Pro Units until Canikon release a 5D Mark III and D800. Full frame sensors, big bodies, pro-grade support networks, lots of huge and fast and expensive lenses. (Canon and Nikon both make even more expensive dSLRs than these, but they are rarely seen, even in the hands of pros).
The presently-small medium-format digital market might grow up and surprise Canikon (I'm looking at you, Pentax 645D), but I'd bet that full-frame dSLRs will remain as the pro camera of choice.
The Amateur Market
Photography is so widespread that almost all of us are photographers. But some people take it more seriously than others.
The serious hobbyist market spans a fast array of niches, but its present and near-future is centred around cameras where you can change lenses. The present amateur dSLR market features some fabulous cameras: the best have low-light performance and image resolution that was the territory of pro cameras 5 years ago, and they add HD movie-capture capability that in some ways matches the best pro cinematography equipment. (In other ways, it's not so good, more on that later).
And yet, they look like a dead end now. First Micro-4/3, and now the Sony NEX series and others are all demonstrating how to put big sensors and interchangeable lenses in very small bodies. Giving up the reflex mirror and box (in exchange for the live rear screen) is probably a benefit for most amateurs, not a problem. And the size! These are some very small cameras. A lot of pros add them to their kit, either as a backup or even as a primary travel-photography tool. They're amazing, they're getting better, and they're popular.
(And as an aside, the Sony NEX had a very modern product-development cycle. The initial reviews described how the horrible menu system made key camera settings nearly unusable. Then a revised firmware came out, and the camera was great.)
Even though I bought (and am very happy with) an old-style dSLR, I think the world is changing, and these small interchangeable lens cameras are the present and near future.
But as for Micro Four Thirds ("u43" from here on in) as a format, I think it's doing great now, but won't do well in the future. The basic problem is that u43 cameras have smaller sensors than their competitors like the Sony NEX (the bigger sensor is an APS-C, same as most non-pro dSLRs), but the bodies and lenses are no smaller. It appears to be affecting low-light performance (not to mention the beloved depth of field) relative to the competition right now, and physics suggests this gap won't close much.
Cameras with interchangeable lenses (both pro and amateur) and big sensors have picked up a new trick in the last few years: video capture. Since an APS-C sensor is nearly the same size as a typical Super 35 film frame, these cameras have a distinctly cinematographic look, optically.
These cameras are not simple video cameras: they are fussy, prone to odd limitations like short recording times (sometimes due to sensor overheating!), limited exposure control, noisy or unsuable autofocus while filming, and a host of other limitations that make both non-hobbyists and pros throw up their hands.
However, these limitations are very close to what film cinematographers have always had to live with, which is why they carefully stage shots, plan and lock in their focus and zoom movements for each shot,
and generally spend a lot of time planning and fussing around to shoot each short scene.
In other words, for the cost of a decent amateur camera, you can produce results that look like they came from a high-end film movie camera. If you have the talent and patience, that is. Also, you'll have a harder time booking Bill Murray and Scarlett Johansson.
Bill and Scarlett aside, this is amazing. I intend to put my camera's video mode to good use.
What should I buy today?
For most people? A phone with a camera in it. Don't forget: your camera should be small enough that you carry it anywhere.
For parents and non-hobbyists who need something more than a phone-cam? All the current name-brand cameras are great. Really. Either buy a camera that fits in your pocket, or buy the smallest camera whose optical zoom is long enough to work for your child's sport.
For serious amateurs? You know better than me, but the Sony NEX-3 is an obvious, $400-when-on-sale sweet spot for the photographer who doesn't already have lenses. For my part, I'm very happy with my Pentax K-x and a collection of inexpensive lenses.
What will I buy in the future?
I expect cameras to become even more ubiquitous, and continuous, and behold the Looxcie: an ear-mounted bluetooth headset/camera that continuously records what you see. When you see something good, you push the button and it saves the last 30 seconds of video. I think that (or something very similar) will eventually become a very common accessory. And note that it resolutely acts as an accessory to your smartphone, and thus it is attempting to ride the trend rather than fight it.
I expect that cameras, as I said at the start, will become devices where sophistication is removed from the physical imaging system and added to the image processing system.
This is a bit of a lie, since I'm talking about physical imaging systems that haven't even been invented yet, and which will surely depend on the capabilities of the modern world's very sophisticated semiconductor manufacturing processes. But I mean that the fancy and expensive parts of a fancy and expensive camera these days are the sensor and the lenses. And of those two, the sensors are getting better and cheaper very quickly, while lens quality has barely changed in 50 years, and prices are not changing either.
There are some clever, mostly theoretical (for now) ways to work around the need for an expensive lens, and most of those tricks involve throwing computing power at the data the sensor captures. It's the future.