The first thing you should read is the RDF primer [1]. It explains the basic data model that the semantic web uses as building block for anything else.
The gist of it is to provide a way to model data as a graph, and make it almost trivial to exchange data between parties. A simple example is, if you and I want to share, say, user data, and you have a "people" table and I have "Users", and my fields are "first_name" "last_name" and yours "given" and "proper", and so on, how do we match the same user across tables? RDF provides a solution for that.
The "pie in the sky" part is that, from this basic building block, a whole host of other technologies can be built, some of them dealing with artificial intelligence, machines being able to _infer_ more information from what you have already said [2]. Inferences can work best when there's more data, that's were the ability to mix your data with third party databases comes handy.
Solid adds even more stuff to the table. In their words: derived from "social linked data", is a proposed set of conventions and tools for building decentralized social applications based on Linked Data principles. I think "decentralized" is the keyword here. I haven't even started looking at it :-)
1: https://www.w3.org/TR/rdf11-primer/
2: https://www.w3.org/standards/semanticweb/inference