It depends on your target audience, and your workflow...

It is in some cases better to WORK in Adobe RGB and then PUBLISH in sRGB if your main target is web-publication. But this also adds two things to your workload:
1) you have to manage saturation clipping yourself
2) you have to remember to ALWAYS convert to sRGB before publishing

If you're shooting normal situation targets (not flowers, not synthetic material commercial product, not very saturated color clothing) it is often enough to keep an eye on what you think could be / become oversaturated areas this way keeps the largest amount of information in the image.

But: My standard advice to avoid over-complicating things (and thereby increasing the risk for later mistakes!) is to try and learn the new local-adjustment gradient- and paint-in effects in ACR and LR. Do this while softproofing the output as last stage in the raw workflow, and you've made sure that as much as possible is processed in linear ProPhoto (internal processing in the ACR and LR engines is ProPhoto gamma 1.0).

Use them as much as you can - to avoid doing to much heavy lifting inside Photoshop later.

I've totally stopped archiving "intermediate" files.
I have raw files, and I have development settings - then all client-oriented material is saved and archived as-is (jpg-high in whatever colorspace the client prefers) in folders named with client, date and ordername.

Yes, I do know what happened in Eivind's case, and it's entirely my fault.

Since he had some trouble using the software (actually - he didn't bother to read the instructions...) we had some contact just after he bought the card. In this mail exchange, he very graciously enclosed some raw images where he could not get the program to work correctly.

I have two very large folders of sample images, one is what I call "client problem images", and one is called "sample images". Most of the sample image folder is made up by images taken by either me (with some 30+ different cameras) or by local photographers that I know personally, and have guided through some color management issues. There are about 6-800 raw files in each folder at the moment.

Eivind's photo showed a very large difference in purple colors, and since I'd at some point along the line had put it in the wrong folder, I filed an used it as a "sample image".

He reacted very aggressively, and even mentioned lawyers in a mail directed towards us. We removed the image instantly, and apologized for inadvertently advertizing his name and company in the EXIF of the image...

His claim was that we were monetizing his name to promote our product, a stance which I personally have trouble understanding - since the shot is in no way representative of his skill or average work, and we didn't use his name or company name to promote the image.

Using someone's image both un-bylined and without their consent is wrong, and since I of course agree with this I immediately removed the image. All of the images in the sample folder should be cleared and confirmed - and that was my mistake.

But to have the energy to write this blog in that aggressive tone is kind of overreacting I think... As was stated several times, very clearly - he was very welcome to return the card for a refund at any point during our exchange. Everyone is of course free to try other solutions, and compare real results. Since I know the relative quality of all other generally available solutions, I can afford to be fairly confident.

What we've designed stands up for a really intense scrutiny both metamerically and from a digital color quality PoV, even though it may not come in the most exclusive packaging you've seen.

Sorry for the long OT response, but I had to react to that link...

Please note:
I have a personal business interest in QPcard, so weigh that in when reading my posts regarding this particular matter.

Quite a few (eight) of the pigment patches are well outside sRGB, and the greenish blue is actually outside Adobe RGB too.

This doesn't really matter either - since colour perception is linear at XYZ basis. A colour that is "50% saturated" is exactly corresponding to the same hue, but "75% saturated".

I wouldn't recommend trying to calibrate anything against reference patches that are less than 25-30% saturation of the target gamut, but anything well over this is more than enough for any real purpose.

You might get differences against purely monochromatic targets, but out in the real world extremely few hues can even approach this purity. Some lasers and some neons do - but they're the only real-world targets I can think of straight up.
...........

Some people (well - me, and a few colleagues in the business) like to have full manual control over gamuts, so ProPhoto > softproofing > manual editing is the preferred way to get absolutely maximum out of the print chain. But if you're slightly less obsessed with colour than us nutters, then Adobe RGB will cover 99.9% of all inkjet output. Some of the newer 9-pigment printers will go outside ARGB (but only in the darker hues!), but they're well down below perceptual threshold differences. In double-blind tests we've seen no actual output difference.

ProPhoto is good when you have many intermediate steps pulling first one way and then another, but ONLY when the entire chain is 16-bit. When some part of the chain is 8-bit depth, then ProPhot actually hurts the image quite bad. Posterization is amplified by about 300% comparing sRGB or normal inkjet 8-bit to ProPhoto 8-bit.

Hi all...
No, you generally don't need a different setting for different output spaces.

How it works is like this:
When set at "Large", there's no modifications at all. You get a profile linear all the way out to ProPhoto limits.
When set at "Medium" the saturation softclipping starts at about 80% of Adobe RGB. That's a very saturated colour anyway. This is what I normally recommend.
The "Small" setting is very usable if you KNOW that you're going to print images on a laser color printer or something like that. Also for web-only images, but then you probably want to bump saturation a bit in ACR or lightroom to get some overall "punch".

ACR does not "handle" gamut clipping in any way. It's a purely textbook "relative intent" clipping behavior.

If you want a better handling of clipping (and sRGB...) then the icc-consortium have a beta sRGB icc v4 perceptual profile on their site. But that's rather cumbersome...
ACR>export AdobeRGB>open in Photoshop>convert to sRGBv4 perceptual>convert to sRGBv2 standard IEC profile>save file.

You don't want to publish images with the v4 profile, since most receivers won't be able to interpret it correctly.