Depending on how they structured their deal, it may very well be possible that Springer knows about this and is allowing it. There are quite a few other Springer books available on authors' websites, for example The Elements of Statistical Learning (http://www-stat.stanford.edu/~tibs/ElemStatLearn/)

Except the book isn't available on his website. His website has link to a warez site with his book on it. So whatever deal he has with Springer it doesn't seem to allow him to host the book on his site.

He does host a simplified version of the book however, 'EZ version' as he calls it. I'm sure the 'steal this' full version comment is just tongue in cheek and/or sarcastic. Perhaps even a veiled prompt at Springer to send their lawyers after LibGen, who knows, perhaps it hasn't happened as fast as he wished for, thus the comment. Or maybe it just saves him some bandwidth if people get it somewhere else :) So many possibilities.

Springer's books cost so much that I'm sure even Springer doesn't expect to sell them directly. Their money comes from academic subscriptions that they will get regardless of how their readers get the content. This is why I think they are so fond of the golden model -- they just want tonnes of money for their brand. Elsevier is a nest of sociopaths in comparison.