With the release of UML 2.0 into the wild, it was inevitable that Ambler would update his Elements of UML Style book. Like its predecessor, this little book is more a companion than a teacher. You will not learn how to do UML modeling from this guide--that is not its intent--but you will learn how to make your UML models more readable and consistent. Ironically, books that do intend to teach you UML modeling do not talk about "goodness" or accessibility of the model's representation. This guide brings that balance to your use of UML.
There are many "gems" sleeping in his book. In his section on the Sequence diagram Ambler recommends putting the message text near the receiver side of the message arrow. This is not subtle: it makes the model significantly easier to read--especially when evolving the diagram in real-time on a white board. "Prefer Names over Types for Parameters" is another, and especially true when modeling from an analysis mindset rather than describing design. I could go on, but that would be redundant. Get the book. Read it. Consider his recommendations, and feel free to go in another direction if you can justify that direction is an improvement in clarity and communication.
Not everything Ambler says should be taken as "the way", of course. As an experienced modeler I have certain style preferences to which I gravitate, and which violate some of Ambler's precepts. In use case diagrams, for example, I always denote an arrowhead on the actor-to-use case association. I also note specific stereotypes on the actors of a use case diagram because not all actors in a system are equal in importance in the project lifecycle. But I follow, and have independently adopted, many of Ambler's recommendations. Indeed, I know that many of his recommendations are derived from the common usage of the modeling community, plus his own unique additions.
If you are just starting out with UML diagramming, this little guide will help you adopt some consistency as you navigate the new landscape of 13 diagrams in UML 2.0. Not all of those diagrams are equally important, either, and I was very pleased to see Ambler clearly stating his reservations about the Composite Structure diagram. Many of us wonder about the value of this offering in UML when other issues such as data modeling and UI modeling have been so prominently absent since UML's inception.