> The least this would do is avoid a number of unnecessary calls to > instanceof.
You're on the right track, and maybe that "switch" impl would be an 80/20 win over the cost of the Right Way. But that huge if-instance case is exactly what the Visitor pattern was created to solve. I haven't spent enough time with the Fop codebase to speak to the gains of such a thing, but I'd bet RTFHandler isn't the only one that does this. A braindead `grep instanceof | wc -l` search shows 122 classes out of the 640 classes use instanceof, which in my world makes them at least worth examination. Another "implied" win to the Visitor interface is if a new FONode type gets added, it forces the change to ripple (unless the impl is using the DefaultVisitor) to all those who would need to handle the new functionality. -- /v\atthew
