I'm guessing it's either because of scope or because generational collectors are a type of copying collector. Instead of having just from/to spaces that the objects ping-pong between as they fill up, a generational collector will copy objects from one heap generation to the next as they fill. Once the oldest generation (tenured space) reaches capacity it cannot trickle down the live objects to another generation so some other method, like a mark+sweep+compacting method, is used for collecting it. You could probably have a half space tenured generation but half space collectors are not really space-efficient without resorting to a bunch of virtual memory mapping trickery, and the tenured space tends to be large.
It isn't mentioned because the author chose not to include it.
Whether that's because this isn't meant to be a complete survey of the field and the author didn't think it fit the introductory nature of the piece, or because it didn't fit within the time limits of the talk this article is based on, or because Ruby doesn't do generational optimizations and the author saw no reason to wander down that path, isn't a question that expertise in gc algorithms gives much insight into.
