That's the question Rob and Joanna asked here. Here's how I see it (and by that I mean, here's how I think it should be, not how it currently is in the jurisprudence!).
There are three elements to determining whether a measure is discriminatory.
1. The Effect of the Measure
There are a number of terms that can be used here: Disparate impact, discriminatory effect, or detrimental impact (the AB's latest term for it, as I read it). What are we looking for in terms of the effect? In my view, it is the degree of disparate impact.
When you have a de jure discriminatory measure, this is easy: 100% of imported goods get the worse treatment, and 100% of domestic goods get the better treatment. The degree of disparate impact is as high as it can be.
With de facto discriminatory treatment, it's more complicated. I would say that you are looking for an appreciable disparate impact on imported goods. If 60% of imported goods get the worse treatment, and 60% of domestic goods get the better treatment, that would probably be enough. The higher the degree of disparate impact, the higher the likelihood of finding discrimination.
Cases like COOL and Korea - Beef are a little tricky to fit into this analysis, but if the measure creates incentives to use imported goods, I think it's safe to say that a fairly high degree of disparate impact exists.
2. The Objective Intent of the Measure
Under this element, you infer the intent of the measure from the terms of the measure itself. I've always thought that the "design, architecture, structure" terminology provided a good description of this element. And I think Chile - Alcohol offers a good example of how this approach can be applied to understand the true purpose of a measure.
3. The Subjective Intent of the Measure
This is what legislators, regulators and other government officials were saying in relation to the goals of the measure. A lot of people are nervous about relying on this. I think it should be a relevant factor, although certainly not the deciding factor, and probably not as important as the other two.
To determine whether discrimination exists for a particular measure, you examine each of these elements in relation to the measure, and then weigh and balance them. One part where it gets tricky is that the effect and objective intent elements seem to overlap some times.
So, how do these elements appear in the recent TBT 2.1 cases? Well, I think that the "effect" part correlates somewhat with the "detrimental impact" analysis the AB has been doing. On the other hand, they don't seem to be looking at "subjective intent," although that might in part be due to the way the parties have made their arguments. They have looked at government officials' statements in the past, and it's not inconceivable they would do so again.
The element that's most difficult to judge is "objective intent," which is, arguably (and somewhat confusingly), what is looked at under the "discrimination" element of 2.1. What does the AB think of this element? After Cloves and Tuna, I might have argued that they were applying this along the lines of what I described above. After COOL, though, I'm a lot less sure. The whole analysis (see paras. 341-349) felt very different, perhaps more like what Rob and Joanna characterized as looking at whether public policies are "reasonable or understandable." That's not how I see the "design, architecture, structure" standard, but my sense is that some people do see it that way, with the analysis focused on the means-ends relationship or something similar. There may also be some people who think "reasonableness" is just a good, general standard to apply to the acceptability of domestic measures in relation to trade law. My concern is that many measures will fail this standard simply because governments cannot be that precise in crafting measures. The result will be that measures that are non-protectionist in intent, but badly drafted and ineffective, will be found to be violations of trade rules, and I'm not sure the system can handle that.