Just as an aside, it isn't reverse discrimination. There is no such thing, logically speaking.
Inasmuch as the phrase 'reverse discrimination' more precisely describes a phenomenon than the more general 'discrimination' speakers will employ it.
As "logical" as you approach might be, it functions only by decontextualising discrimination. And by positing various discriminations as abstract and equal, "just as racist|sexist" one blind oneself to any systemic cultural biases which might operate against subaltern groups. It is not as though the prevalence of the biases that exist in a culture represent merely the sum total of individual decisions to be biased in any particular direction, any more than the prevalence of English use in anglophone countries can be explained by the sum total of individual choices made by speakers.
In contradistinction to this neat decontextualised logical analysis, an empirical study should turn up actual systemic discrimination where such discrimination exists. But it might turn up more. In a culture where it has become possible to appreciate systemic bias within that culture, individuals (and here we are more likely dealing with the deliberate and reflective decisions of individual members) may decide to (over)compensate for the systemic biases they perceive to operate. That particular form of discrimination, as distinct from others, is 'reverse discrimination' and one needs to account for the possibility that it is affecting observations.
Thus while it may not make much sense purely in the abstract, in real-word situations it has obviously become necessary to discriminate be 'discrimination' generally and 'reverse discrimination' in particular.