A successful open-source project is also about building a community of third-party contributors. Or at least the possibility of one such community happening. Dual licensing basically kills off that discussion, because in order for dual licensing to work, then contributors need to sign a contributors agreement that licenses their work for whatever purpose the parent company sees fit.
And this is where I've always disagreed with FSF's stance. Many people claim that GPL is inadequate because of the "ASP loophole". But a much bigger loophole is dual licensing. This is because dual licensing happens (and is useful) only when the product is unusable as open-source. For example Gitorious is AGPL licensed and does not need dual licensing because you know, it's meant for people to have an on-premise solution for Git hosting and it's a take it or leave it thing. A Javascript library on the other hand is totally useless as AGPL. Dual licensing is also what made Qt and consequently KDE to be less popular than Gtk+ and Gnome in distributions such as Red Hat's or Ubuntu, or why popular project such as Firefox or LibreOffice are better integrated with Gtk+/Gnome, even though Qt has always been technically superior. Heck, Qt's licensing is why Gtk+/Gnome happened in the first place.
You see, the biggest advantage of open-source, its raison d'être if you will, is the possibility to fork. But if a fork cannot survive because of licensing reasons, maybe because it cannot be applied to the same projects as the original, then that's not really open-source.
Personally I prefer the open core. For example I currently work with Scala and the JVM and I use IntelliJ IDEA as my IDE. It has a community edition with everything I need. Yet I pay for the Ultimate edition anyway. If the community edition gets discontinued, I'll switch in a heartbeat.