The trick seems to be a combination of both. Learning by rote in combination with an effective method seems to be key to strong number sense in kids. Take a look at the Triple Code Model (Dehaene & Cohen) for some more detail.
It's mostly sufficient. /21000/ will not match "http://picolisp.com/21000, which is the first URL in the sequence, but the remaining URLs look like "http://picolisp.com/21000/!start?*Page=+2, so Googlebot will likely only continue to download a single page once it has re-read the robots.txt.
Which is what you deserve for using non-standard URL formats.
No, I'm saying /21000/ will match a path with a directory named /21000 but not a file named /21000.
When I say "non-standard", I am saying am saying that if the website's URLs looked like "/21000/foo" and "/21000/foo?page=2", it would have been easier to craft a "Disallow" rule that would have successfully blocked all of the desired pages.
