(Some caveats:
To talk about the elephant in the room, we both know we are engaging in an old debate about double-standards and 'reverse discrimination'. Let's not pretend it's some novel thing we're proposing.
Also, I don't know about the credibility of the first citation's definition of hate speech, and I think Wikipedia has little credibility (no idea why I'm suddenly seeing so many people citing it on HN) - it might as well have been written by you or me. So I'm not commenting on those.)
Some points:
* There are degrees to threats and danger, of course; communication depends on words, context, emotion, the parties involved, etc. Every insult is not a threat and does not rise to the level of hate speech in a significant way. Perhaps we should use the term 'threatening hate speech', to distinguish it for these purposes.
* There is a clear standard necessary to 'threatening hate speech' that is omitted from the reverse discrimination argument: The vulnerability of the target. To reuse an example I gave in another comment, if a Swede is locked in a room with 30 Norwegians and says, 'Norwegian scum should die', people might back away a bit but it's not much threat. But if a Norwegian says 'Swedish scum should die', that's a threatening and dangerous situation for the Swede. In U.S. law, this standard is expressed as 'protected classes' (afaik). Hate speech is only a double-standard if we omit this factor. The consequences to vulnerable people are very well established and very dangerous, arguably the worst events of human history; we can't just ignore that because we lack the perfect solution. Doing nothing is not the null hypothesis, it's an alternative with clearly worse results.
* Subjective judgments are less desirable than objective measures, but that's life; the world is just too complicated and in almost every case we don't have time for scientific accuracy. It's a theoretical argument to require objectivity; reality, as everyone can't help but observe, requires and uses endless judgment. That doesn't make all judgments good or equal (another theoretical argument). Note, for example, that courts have a complicated system designed to output good subjective judgments by judge and jury; they expect and accept subjectivity. But to the extent a rule requires subjective judgment (which is every rule, as every one has edge cases), I agree that it would be better to do it objectively, we should do that where possible, and that we have to be careful.