Commenting should be how people down vote. You should have to explain why it's not a good idea. Then if others agree with you, your comment will get up voted. Just as my comment is a downvote of github's downvote.
The context of this feature is that users were using comments to react without providing substance. The primary motivation is to reduce the visual clutter of non-substantive comments.
While we can agree about what users should do, users not behaving as we think they should is the whole point here. Removing the -1 reaction won't encourage people to articulate their concerns. Nor will it discourage people from writing comments that explain why something is not a good idea. It simply provides a mechanism for people to register disagreement that is less obtrusive than the mechanism they would otherwise use. A cavalcade of -1's as comments obscures and thereby prevents substantive discussion.
What users wanted was a way to gauge what features are more important, so maintainers could prioritize the important issues to the community. All asked for was to be able to up vote _issues_, and be able to sort. [1]
Instead we got reactions, and it completely missed the mark. They've made it easier to be passive aggressive. Also it lingers on your post, instead of having the unconstructive comment stand on it's own. How does one respond to a down vote, make another comment saying "... hmm?"? It's weird.
Facebook and Twitter provide strong evidence that people will use reactions. I see no reason to expect that user behavior will differ when they want to express negativity. Facebook's decision to not provide this feature has long been rooted on the pretense that people will use it. That is the basis of this thread: that people will use it instead of commenting.
Also, it's really easy to disagree with what I wrote without suggesting I am naive.
I agree with how you want to conduct the PR or RFC process, where it's annoying if people are being negative without providing an alternative. I do however agree with the parent where there are plenty of times when a simple yes/no vote is what you really want. The comments are there for people to provide alternatives still, removing the -1 would just be removing flexibility for a very common case.
Then such a vote should be opt-in - that is, it should only be possible on posts where the writer has explicitly chosen to make this a possibility. That would solve the issue.
While adding another piece of state and another UI control to an interface that, AIUI, did not require any changes to support this feature, which would have delayed shipping and getting this functionality into the hands of their end-users.
But making an opt-in, allow-negative-feedback control is not feedback I think they'd be wise to listen to. Users can edit issues. I'm finding it exhausting just to think through the implications of either allowing editing of the flag, or deciding that users have to get it right the first time. I can't imagine I'd find it easier to use - in fact I am sure I would be intimidated, plain and simple, by the fear of who-knows-what happening if I did the wrong thing.
(...but then, that's the most git-like thing I can imagine doing! something-like-a-VCS that you can build entire new companies' worth of brand-new development practices around and still be utterly impenetrable without first unlearning 30 years of perfectly functional version-control experience and collective wisdom. I can't express how glad I am that I don't make things for developers. Truth is, most user bases are a joy to work with. Developers... insist on dysfunction, it's non-negotiable. I think it's part of the whole, "we still don't know how to do good software engineering" schtick. It wouldn't matter if it were a solved problem to the point of a mathematical proof, nobody would allow it anywhere near their mysterious uncharted edge-blurring code artisanship. We really, really suck, both professionally, and at being adults. For the most part, anyway.)
