Gleb Mazovetskiy

I think I see a bug with that query. Just to confirm, what happens if you set `OrderQuery.wrap_top_level_or = false`?

Thanks, I've opened a separate issue about this bug

Sounds good to me!

I don't think the minifier solution is viable, because concatenating CSS may result in different selector priorities? WRT feature creep, this actually seems like the most common use case.

@david-trejo-ck Hi! I ended up using uncss https://github.com/giakki/uncss

uncss does not use phantomjs anymore, the current version uses jsdom. I haven't checked the coverage numbers in Chrome devtools yet, but visually looks ok

`@font-face` worked for me but I filter them out with [filter-css](https://github.com/bezoerb/filter-css) anyway because custom fonts are not critical (even icon fonts)

Ouch, I hadn't realized WeakMap/Set are not iterable. I agree that this means `eqv` and `equal` versions are basically impossible. There is a stage 1 [tc39 proposal for weak references](https://github.com/tc39/proposal-weakrefs/)...

The weak references TC39 proposal is now stage 2: https://github.com/tc39/proposal-weakrefs, meaning "The committee expects the feature to be developed and eventually included in the standard".

> If my JS memory serves correctly, smaller prototype chains are desirable. That's true, however, this would *not* increase the length of the prototype chain, as they would all inherit...