Generic Query Conventions Need Defining

[adamnovak]

Right now, we have a lot of SearchWhateverRequest objects, each of which contains some number of search parameters.

The enrage idea seems to be that most parameters take arrays of acceptable values (but interpret the empty array as any value being acceptable), and that results are supposed to satisfy all of these constraints. However, this isn't actually stated anywhere as a principle of the APIs' design, and many parameters of query objects do not follow this convention (for example, you can only supply a single assemblyId when issuing a SearchReferenceSetsRequest). We should figure out when the array semantics are appropriate and when the single-value semantics are appropriate, and produce a design document that tells us how to add any particular query constraint in the future.

This is related to #247 and #253 (as they are addressing the issue of standardization for some fields), as well as #256 (as the design documentation would probably go in whatever that produces). It also touches on some line note discussions we were having on #250 with @fnothaft, @jeromekelleher, and @pgrosu.

[jeromekelleher]

Definitely +1 on this idea. Hopefully this will be clearer when we resolve #247.