Hi Troy,
Would like to "set and forget" my colour management
Approximately one half of humankind would like to achieve that. The other half don't work with images at all
I'm pretty sure the monitor profile rendering intent as "perceptual" is correct
The perceptual rendering intent is the usual option for most monitors. If your monitor is a good/modern one, it should be able to represent a much wider gamut than sRGB, so you may prefer relative colorimetric. For example, the DELL U2711 where I'm writing this post is able to represent most of the Adobe RGB(1998) space, so relative colorimetric is a more appropriate option since it matches white points without rescaling colors, and it is quite unlikely to have out-of-gamut colors with this monitor. However, the actual difference between both rendering intents is very small, if even noticeable.
Default profiles I've got set as ProPhoto RGB and Grayscale Gamma 2.2.
A good choice. However, bear in mind the following:
I use ProPhoto RGB in PhotoShop because after some research, it most closely matched the sensor of my 5DII.
Note that your raw images are not tied to---should not be tied to---any color profile. They are actually linear numeric data, not representable images. If there is some connection between the raw data and a color profile associated with a nonlinear device, then the data are not actually raw, but precooked. The raw data pertain to the canonical RGB space, and only the chromatic response curves of your camera's sensor (and its color filter array) should be used to describe how colors have been recorded and stored in the raw data. Note that no device-specific color profile, neither any device independent profile such as ProPhoto, CIE RGB, Apple RGB, Adobe RGB, etc., can be used to describe the raw data (the main reason being that all of those profiles describe nonlinear color spaces).
Once you have stretched your image, you can decide to work in a wide gamut space that allows you to represent more color saturation and contrast without forcing the numbers too much. This is the reason why people tend to prefer working in ProPhoto, Adobe RGB, and other large spaces. If you're comfortable working in the ProPhoto RGB space, then it is a good choice. One can also decide to work in the sRGB space, which is the default space in PixInsight. sRGB is the standard color space of the WWW, which defines how most people will see our images. Working directly in sRGB has the advantage that it helps us to avoid the danger of too fancy assumptions regarding color saturation.
Again, the practical differences are not significant because you should always use quantitative analysis to discover and prevent potential problems such as histogram clippings, oversaturation, chrominance noise, etc. Relying on color management to make image processing decisions is almost always wrong.
I convert/transform to sRGB (relative colormetric, black point compensation) before posting to web.
Perfect.
If there's no embedded profile, in PS I'd assign the default profile.
Unless you are *completely* sure that your untagged images can be described by your default profile, this policy is dangerous because there's nothing worse in the world of color management than assigning an incorrect profile. It is much better leaving the image untagged. With an untagged image we can always ask what profile should be assigned, but an incorrect assignment represents a terminal---and wrong---decision.
Colour proofing. I've never had a need or used this in PS or LR. What's the common practice here?
Color proofing, aka
soft proofing, allows you to simulate how an image would be seen if represented by a device different from your monitor. For example, suppose you are working in the ProPhoto RGB color space (which is either your default working profile if your image is untagged, or the profile assigned to your image). Your monitor has the Adobe RGB(1998) color profile assigned. What would your image look like on a low-end (sRGB) monitor? How can I detect out-of-gamut pixels when the image is represented on a sRGB monitor? Both questions can be answered with the color proofing and gamut check features, respectively.
The accuracy of the simulation will depend on the quality of the involved profiles. Generally, proofing for sRGB is very accurate because we work on theoretical color spaces. Proofing for printers is much more difficult because properly characterizing a specific combination of printer, ink (or toner) and paper can be hard.
Just to confirm, is it best to leave colour proofing and out of gamut display off by default?
You should leave both options disabled by default. Proofing is normally enabled only when one wants to check the appearance of the image on a particular device.
And finally, the global options. I seem to remember Juan also recommending leaving the ICC profile embedding off by default. Not sure of the reasons why? I've always embedded them in PS.
I probably said that years ago, when browser color management implementations were rather poor. Today we have good implementations at least in Firefox and Safari. I don't talk about Internet Explorer because I don't use such thing. Currently I'd embed the sRGB profile in all images intended to be seen on the WWW (after converting them to sRGB if necessary). IE may be inaccurate for color reproduction (I really don't know), but then IE has caused lots of problems since I can remember (and I can remember long time ago...) so no surprise. Firefox, Safari and probably Chrome are good color-managed applications so I think that embedding the sRGB color profile is currently safe. I wouldn't assume that the average user can see images in a wider gamut space; sRGB is still the only safe option in my opinion.
