If anything in this thread doesn't make sense it's your earlier comment "you can just about forget soft proofing".
The quote is incomplete. Just forget about soft proofing,
using sRGB to Blurb. It's a complete waste of time. It shows you nothing useful.
If Blurb ask for sRGB images, what reason is there not to soft proof to that colourspace ?
The idea behind soft proofing is to simulate on screen, the output process before you output the document. You are sending your data to a CMYK device. You don't have a profile for that device in this case, the Blurb CMYK profile doesn't define their output(s). Using sRGB is even farther from the truth. Viewing a sRGB soft proof going to a CMYK device doesn't tell you
anything about that output. And as I said, any soft proof that voids the use of the simulation of ink and paper is an incorrect soft proof unless you've properly calibrated your display for that by controlling the contrast ratio (few displays can do this).
In LR, you're viewing in an RGB working space
anyway, the idea to switch to sRGB somehow tells you something useful in it's preview based on a CMYK process is about as ineffective as loading an Epson 3880 Luster profile when you're printing to a Canon ipf5000 on their glossy paper. Why would you ask LR or Photoshop to show you a complete lie in the preview of a soft proof that has no bearing on the output?
You pick the output profile you will use to convert that data going to the output device for soft proofing. Period. You have to also examine the effect of the Rendering Intent and the paper and ink simulation. That is how you view, on-screen, the image as it will appear (as close as our technology allows) on the output device. The sRGB color space has no role here (even on the Web outside
your system).
If you want to get picky, you should be aware that by the time the CMYK conversion process is to start, sRGB (any RGB) is out of the picture, the data will be in Lab. You could just have picked ProPhoto or ColorMatch RGB and the '
soft proof' would be just as incorrect, a lie about the output as if you picked sRGB.
According to blurb, sRGB is the "preferred" color space. The implication is that other color spaces will be accepted but are not ideal. If we Send Book to Blurb via LR, I suppose there is no choice but to use sRGB. But what happens if a book is submitted directly to blurb outside of LR using - say- Adobe RGB? Does blurb convert to sRGB? Does it even know that the images have been sent with Adobe?
In LR, you're exporting sRGB to Blurb. FWIW, in Aperture it's Adobe RGB (1998) going to their CMYK printers. IF they tell you they can take any RGB working space, it means their front end is smart enough
not to assume sRGB (that's good) and they will convert to the output color space from any tagged RGB space. Unlike many silly RGB lab's out there, they will allow you to send them something other than sRGB. Blurb doesn't, as these other labs do, assume and demand sRGB. But when rubber hits the road, whatever RGB working space you provide will play a small role, mainly in terms of out of gamut colors in the source that could be mapped to the destination (yet again, you have no idea what rendering intent or in this discussion profile will be used to end up in CMYK).
Bottom line:
IF you have a service provider that demands sRGB, soft proofing is hardly useful assuming they give you a generic CMYK profile that doesn't define the output fully. Soft proofing to a CMYK device using an RGB working space based on a CRT is just a waste of time, it shows you nothing that has any bearing on the output in terms of a soft proof. Might as well just view the image in it's current RGB working space.
If you have service provider that demands sRGB and a profile t
hey will use for the actual output, and they let you use it fully, that is now very useful as you control the process and presumably (short of using a device link which seems unnecessarily), they will send the CMYK numbers as is, to the press. The data you converted and saw is the data sent to the printer.
For this to work, the display calibration and profile and the print viewing conditions have to be configured correctly. It's 'diffcult' enough doing this to a high degree with the correct output profile. Using a profile that has no bearing on the output and thus the soft proof is just a soft proof that has nothing useful to show you.