ALNP has been used with SPDY for a while now. It one of the nice improvements that fell out of testing/iterating SPDY in public. The NPN approach was a bad idea since the client drove what got picked (With NPN, Server tells the client what other protocols it supports in the ServerHello and the client picks whatever it wants. ALNP reverses that.)
From TFA: "However, there is also some sadness in this announcement. As we transition to DataStax, we will find little time to contribute to Titan and interact with the Titan community. We will miss that and hope that it will be carried forward."
It's a bit hard to elaborate. You have to rember that a classification problem may be generalized to more than just two dimensions, in fact it can be generalized to however many dimensions are necessary to describe the data. In the general multidimensional cases linearly separable relates to the number and type of hyperplanes (subspaces) necessary to correctly classify the data.
In this simple 2-d case the subspace is a line, so this example is linearly separable. In the 3rd dimensional case linearly separable refers to classification using a single 2-dimensional plane. For 4 dimensions with the data, you need a single 3-dimensional hyperplane for it to be linearly separable, and so on.