> But in practice the pull parser still iterates through each branch in
> the parser, but might only return certain parts.
Actually, that's not an assumption I would make. At the point the start-tag
of an element is encountered and the parser is directed to skip it, I
believe all the parser essentially has to do is pattern match the text for
start- and end-tags of same-named elements. When the end-tag count matches
the start-tag count, you start generating events once more.
There's no parsing going on of the skipped content, so it's not iteration.
It's grepping.
