This is a confusing issue, I'll give you that. I think there is learning what is better (Best Practices, if you will) and then there is cost/time/benefit analysis (is it worth the extra hassle?)
People can be divided into two basic camps: those that get educated and those that won't/don't. If you don't get educated you can't really move to the next step of deciding if it is worth it.
People who move to the next level will then be divided into at least two camps: Those who feel the extra time/care
is worth it and those
that don't. Part of that may be dependent upon the equipment that they have.
For example, I think everybody would agree that it is better to keep as much information in the digital file as possible and converting to a lossier format copy only when needed for a specific purpose. That means shooting in RAW in the camera and using ProPhoto RGB in your RAW converter and Photoshop workflows. The reason is pretty simple (i think): ProPhoto RGB is 16-bit, while Adobe RGB and sRGB are only 8-bit. From the link Mike shared above:
Using ProPhoto space is easy. In your raw converter, either output your files into Photoshop tagged with the camera's profile if you're using Capture One, or tagged with in the ProPhoto colour space if you're using Camera Raw. (ProPhoto RGB is Camera Raw's native colour space). When using other raw converters it will depend at the converter's capabilities.
Now why would you want to use a huge color space when your camera limits you to a smaller one? I think it comes down to remembering that computers (software and hardware) treat everything like a number. Imagine that you have a calculator that can only DISPLAY whole numbers. If you input a problem like 9 divided by 2, we know the answer is 4.5, but your calculator has to choose between 4 and 5 for the answer. It doesn't do this randomly, but follows a set of algorithms (in this case probably rounding up to 5). The point is that either answer is not precisely the one you wanted/needed. The same is true with colors. If your software is trying to determine how to render a particular pixel, it will be doing it with an algorithm that takes into account the color information of the pixels around it. If you are working in ProPhoto RGB space in your editing program, you have (in effect) more decimal places to work with and thus a more precise result in the math that determines that pixel's color. The more accurate that is, the more accurate the output will be when it "rounds down/up" to a smaller color space.
While it would be NICE if you could see these differences on your screen, just because you can't see them on your screen does not mean they are useless. It means that you will have a more precise "mathematical definition" of your digital image WHEN THE TIME COMES to round down to sRGB (for example). It is then that you might see a difference, but that is when it counts.
Now back to the original question of Adobe RGB vs sRGB:
Using the same logic as that above, it would be preferable to have as much color information as possible... so theoretically Adobe RGB would be the best choice for in-camera. Again, it would be great to be able to SEE that
difference in your correction process. But I don't think you will/can unless you get a "wide gamut" monitor (and calibrate it properly)
AND your operating system (and software application) supports that "wide gamut".
From what I've read, it sounds like this is where Vista is better than XP (with XP you are out of luck on wide gamut) and where Windows 7 will be even better.
As far as workflow is concerned (Camera RAW/Photoshop), it seems that a lot of photographers "live in" ProPhoto RGB in 16 bit mode and convert to AdobeRGB 8 bit when sending to a pro printer and, of course, convert to sRGB for web use.
Interestingly, new browsers like Firefox 3.5
will take into account an embedded color profile (which I think means that if the user is on an operating system that supports it, and their monitor supports it, they will see a better image than Joe User on XP with IE. This may not be important to most people, but I wonder if there are sophisticated Photo Editors/stock photography buyers out there that will be looking for images to purchase using such a set up. If so, this means that a photographer that takes the time to incorporate this into their workflow will have noticably better looking images (to user with that set-up combination) and therefore
an edge. It may even become a way for buyers to further separate the Men from the Boys (pardon my sexist language).
The downside of this is that a web image with an embedded color profile is going to be a larger file size (and thus a slower download) than one without it.
PS... this is a great explanation of Color Spaces and Calibration, for those who are (still?) confused:
http://digiichi.pentax.jp/english/tech/vol_39.html