Basic questions about the CM handling of Photoshop

Question

I have some basic questions how Potoshop handles colour management. Most probably this has been discussed many times before, but frankly, the manual is pretty poorly written on this subject. I do understand the principle of colour management pretty well, but am not sure whether I understand what Photoshop is doing. So could somebody please comment whether my reasoning is correct or not.
Let's assume my monitor is calibrated and its profile loaded in OSX's monitor preference panel. Let's further assume I'm working strictly in RGB.
Is the following correct:
• When I open a picture which has an embedded profile, then Photoshop first calculates the "true" values from the embedded profile (whatever this means) and from this point the screen representation based on the profile which I defined in the system pref pane. So the screen representation remains always the same, regardless of the embedded profile (besides some possible conversion artefacts, especially on the edges of the colour space).
• The same holds true in the opposite direction, so when I work on that picture on screen, then no matter which RGB space I pick as my working space, the result will always be exactly the same (except those aforementioned artefacts).
• Hence it doesn't matter which RGB colour space I choose as my working space, it could be sRGB, Adobe RGB, "my own wild profile", whatever, the results are always the same. Of course the RGB values will differ, but this will be compensated by the embedded profile.
• Yet the theoretically best working space is the colour space of my monitor, because this minimizes possible artefacts.
• When I "convert" the picture from one profile into an other, then not only that new profile is assigned and embedded, but also the RGB values are recalculated accordingly, so that the result will remain the same. For example, I could first "convert" a picture from the camera space to my monitor space, then work on it, and finally "convert" it to sRGB before uploading it to a web site.
• When I "assign" a new profile, then a new profile is simply assigned and embedded but the RGB values are not adjusted. So the RGB values remain the same, but because a different profile is embedded, the picture will actually be changed. Since Photoshop will recalculate the screen representation based on that new profile (and the monitor profile), I will see this change on the screen. (This action doesn't make much sense, except maybe to simulate the possible outcome on a system which misinterprets the profile, e.g. assumes sRGB.)
• And finally, when I leave the picture untouched but change the monitor profile in the system pref pane, then of course the picture will not change, but its screen representation will change, because Photoshop uses that profile to calculate the screen representation. (This action makes even less sense, of course.)
Have I understood? Then I have two final questions:
• When I open a RGB file without any profile in it, what does Photoshop assume? How does it calculate the screen display?  What does it do if I "convert" such a picture or "assign" a profile?
• If I select in the colour settings as working color space the monitor space, then the colour managment rule for RGB changes to "Colour management off" and can't be changed to a different value. I don't understand this.
Thanks for your patience.

Answer 1

• Yet the theoretically best working space is the colour space of my monitor, because this minimizes possible artefacts.
No.
(Edit: Any one screen usually has to be considered incidental, changing and transient and is not the target-medium, so it makes little sense to work based on its specific limitations.)
• When I open a RGB file without any profile in it, what does Photoshop assume? How does it calculate the screen display? What does it do if I "convert" such a picture or "assign" a profile?
If one opens an unprofiled file, Photoshop treats it as if it had the profile selected under RGB-Working-Space embedded and calculates display and conversion on that basis.
(Edit: And that’s the danger of unprofiled files – they would look different if someone edited them who used a different Working Space.)
Assigning such a file a profile is no different from assigning a profiled image a different profile – the RGB-values remain unchanged, the »key« for interpreting them changes and subsequently it will be displayed differently.