We (monkeys) have agreed that we don't delete data in our databases (to preserve history). We only mark it as deleted (or obsolete). So when we publish our database as open data, everyone (with a little bit of skill) can look up deleted data.
This is such a big topic, it is worth a discussion thread alone. It leads to the point that "facts are not copyright-able". The OSM view on this topic is: "If it is there on the ground and visible to the public, no one can forbid us to map it." (But you can always mark the POI private access only.)
So if we have a map with e.g. all fair-trade shops, we cannot prevent users to add a specific shop, if it fits in the category and exists in reality (when we use an open system).
You're right. You cannot delete information from the Internet - regardless from the license! Even if you publish something with (c) You, you CANNOT prevent the 'bad guys' from making a copy and release it on Wikileaks.
I know it is hard to digest - but with publishing data on the Internet (regardless of the license), you enter the Post-Privacy Era. There is no possibility to take it back. You can never be sure that a search engine or the Internet Archive hasn't indexed the content yet and made a backup somewhere you have no legal access to.
This is my personal opinion, very black-and-white. It depends which kind of data someone wants to protect. I would agree to have different licenses to different kinds of data. E.g. some CC-BY for images, PD for geodata. How to handle contact data is a BIG question.
I think we should start a licence working group