Hi Scott. I'm not questioning (at the moment, anyway) how mmap() has been implemented by the OS. All I want is to force Sqlite to use the :memory: codepath. I'm happy to provide an extra pragma to say "switch to :memory: codepath". Is there some technical barrier to having such a thing? The :memory: codepath already exists. mmap() has already provided a pointer, the same as malloc() does. All I want to do (at the moment) is marry up to the two. If there's no existing pragma to do that, and there is no technical barrier, what would be involved in making such a change? ie approx how many lines of code? Thanks. Paul.