Semantic web has to do with structured data where the information on a page isn't just a combination of words that are put into an algorithm but is contextually relevant data. This way Google can draw correlations between two "entities" and combine them despite the lack of an overt connection (a shared keyword).
Basically it's about Google knowing the context of a page or a query and providing targeted results that may not share the keyword but are intimately tied with the topic of the query.
After writing me own definition, I realized that I didn't answer the original question (I even asked a similar question about a month ago but got 2 reponses. not much of a discussion)
Personally I don't think too many of the recent techniques will really change, they will just be more targeted. With semantic web what's the point in gaining links from non-related websites? In fact I bet that would hurt more with full semantic search than it does now but on page optimization wouldn't really change that much. You'd still target a topic, you would probably have to add context (im not clear on how they intend to implement semantic search. Will that fall on the optimizers or be baked into the search engine?)
For now though, the only thing I can imagine that would prepare you for semantic search would be to implement schema.org and structured data. Use the data highlighter in GWMT or set up authorship/publisher...ship... I think that's a word. Those are the only current "semantic" signals right now and they're effects are hotly debated.
This at least is my view on semantic search and how it will change things. I'm very interested in this topic and if any of you disagree or would like to add/append to my opinion I would love to hear it. There's still a lot to learn on this topic.