So, how much memory would a real-world parser actually consume given this file? I'd try it, but I had to RMA my workstation's motherboard yesterday, leaving me with a machine that only has 3GB, which is the obvious minimum for a full expansion. But I could imagine an XML parser might use UCS-2 internally, inflating this to 6GB. Or, some parsers might be clever and not attempt a full expansion.
So you're asking how much memory this resoource exhaustion attack consumes when you run it.
Each line in the WP examaple amplifies by a factor of 10. It has 9 lines. It's 10e9. That's a billion times 3, which is just enough to 32-bit virtual memory space in common operating systems.
Of course the XML implementation could be smart and short-circuit this while preserving the semantics.
