What are the differentiating features between Highland.js, Kefir.js, and Rx.js?

Since subjective answers are to be avoided on SO, I'm specifically asking: What are the functional and/or performance differences between these three functional/reactive libraries that would make me choose one of them over the others? I have eliminated BaconJS because of its dependency on jQuery. I'm new to the JS streams, and primarily interested in starting to familiarize myself with them first in NodeJS, and later in AngularJS context. If you have selected one of these over the others, what criteria lead to the choice?

Everything below is pretty subjective:)

RxJS

Pros:

• Documentation and big community.
• Shared interface with Rx* family. It's a big advantage because you can adopt samples from RxJava community.
• A lot of primitives in the library and some samples how to use them
• back-pressure
• it is a kind of etalon for other libraries

Cons:

• Sometimes it looks overcomplicated
• Implemented with "Java" in mind

Bacon.js

Pros

• Really nice documentation with some advanced samples(snake game, implementation for example)
• a lot of primitives
• Implemented with JS in mind
• I found it really easy to start

Cons - it doesn't implement back-pressure which makes it silly on backend

Highland.js

Pros:

• it's intended to work with node.js(or browserify) environment which is really nice
• no alien concepts(from JavaScript/Node perspective)
• back-pressure
• the implementation is on top of Node streams which is really great idea
• small footprint

Cons:

• Silly documentation and no examples. Actually every time I'm trying to adopt samples from Rx or Bacon
• absence of some primitives. Actually you can implement most of them by yourself already implemented but it's painful(see previous point)

Summary: because absence of samples you will learn them all:) IMHO, the most perspective one is Highland.js. But RxJs is much more mature

Tip: If you need to ask something on StackOverflow use RxJS dialect and translate it to youth