> Separating the index and the data, even on different storage devices and channels, provides almost no utility. I was thinking there could be some utility, but it would require care. If a fast, volatile device was claimed to be where an index is kept, it could be rebuilt upon opening a DB that claims to have such an index. [a] Then a speed advantage might be obtained relative to what page caching does already. This would require feature change to SQLite to be reliable. [a. Alternatively, timestamping or hash-keeping could be used to verify that a persistent index still corresponds to the table it is supposed to index into. Messy to be sure. ]