I recently came to know that there is a notion called "information-theoretic private information retrieval".
The assumption is that multiple non-colluding servers have same data set and a client wants to retrieve a specific element (or multiple elements) in the data set by taking location(s) or keyword(s) as input without revealing the client's query to the servers.
I have some questions about this type of setting.
Who does guarantee that those multiple servers have same data set at a time?
Is this setting practically used by some companies?