The creation and growth of the IDP field reflects the historical and sociological consequences of methodological and conceptual biases rather than disclosing fundamentally new properties of proteins. All proteins exist as an ensemble of interconverting conformers, they only differ in the number of global energetic minima, and the relative stability of the minima. 

Prof. Colonna asserts that, "...one thing seems certain, the conformers represent a population of molecules in rapid inter-conversion and among them there should be that small sub-population devoted to biological function..." but even this is a foregone conclusion. The implied picture is that there is a global attractor (native minimum) in the conformational energy landscape, but it's not very well populated until, say, a binding partner shifts the dominant populations towards said attractor. This is still a too-narrow picture! A number of putative or established structure-function relationships explicitly require a very large number of disordered conformations. Some of them are annotated in the DisProt database, linked below. Such 'true disorder' structure-activity relationships include fly-casting, entropic clocks, entropic springs, and inter-domain linkers. So I strongly contest the universality of the 'functional conformer hypothesis' when explaining protein structure-function.

The relationships between primary structure (in terms of amino acid composition) and 'foldability' have been recently greatly developed by Rohit Pappu and coworkers. Many assumptions about the basic behavior of so-called hydrophilic amino acids have been invalidated with a first-principles, polymer physics approach. For example, the intrinsic disorder of polyglutamine is apparently a result of it being poorly solvated by water -- even this 'polar' amino acid has a substantially 'hydrophobic' character in the polymeric context.

The problem of characterizing protein dynamics is a large one, almost as limited by our imaginations as our methodological prowess. There is a large industry in the simulation of protein dynamics, but a relatively obscure research area in the interpretation (modeling) of the resulting data. Almost all native state fluctuations are modeled with a fairly narrow set of formalisms. For example, I've been interested in protein dynamics for around 15 years and I only 'discovered' the theory of 'discrete breathers' (as expounded by Francesco Piazza and coworkers) in the last few months. The only thing I'm sure of is that methodological biases can further entrench our conceptual ones. Every successful characterization of dynamics using a particular approach is taken as an indication that the methodology itself is 'successful.' The ability to crystallize globular, ordered proteins does not actually substantiate the universal claim that "crystallography can determine protein structures" because there are plenty of proteins that cannot crystallize (or that crystallize into biologically irrelevant states.) The ability to characterize two- or three-state conformational exchange with a CPMG experiment does not actually substantiate the universal claim that "protein NMR can characterize millisecond-timescale motions" because there are plenty of plausible millisecond-timescale processes that are aperiodic (and that cannot be characterized with most current experimental methodologies.) At this point in time, we need all of the methodologies, experimental and theoretical, to be improved. We need to build apparatuses to characterize phenomena, rather than searching for phenomena to analyze with our pre-built apparatuses. 

http://www.disprot.org/view_function_subclass.php

Ryan put it in a brilliant way. I will try to give some historical (but layman) reasoning. Most proteins studied in the early days of crystallography were enzymes where substrates diffuse to the active site. Enzymes have a more or less stable core hence they can be purified, mutated, NMRed, crystallized, inhibited and studied in depth. Successful design of drugs against enzyme targets have solidified this thought that proteins should have a some sort of structure to function.

Most protein researchers have been bred (myself included) to think that all proteins fold into helix, strand and loops is some way. Not many have heard the term "fly-casting" in relation to proteins (enzymes do not go hunting for substrates). Intrinsically disordered proteins (IDPs) are difficult to study by the present methods. The normal mode of solve structures, then mutate residues to verify functionality does not apply here. Very few IDPs can be assessed for function in vitro. You get lucky if a few residues dock into some other protein (migfilin to filamin in my case). We then fall into the conformation selection bandwagon as Ryan argues above. We have not thought, what, if a protein needs 10-20 or more ( or no) structures to function?

Molecular biologists and crystallographers go to great lengths to coax crystals out of proteins (GPCRs come to mind). These do help, and journals love them. But what about those parts missing density, mutate them and you have change in function. But the structure always wins in the impact factor battle as there is a certainty about it. Confirmation bias then takes over (mutants that cannot be explained by the structure will stay in the lab notebooks for ever).

My point is that the literature gets populated by what can be done given the experimental limitations. SAXS (Small angle X-ray scattering) might be a very useful thing to study IDPs. I smell a Nobel for this area in the future.