On 2/6/16, Bernard McNeill <bm.email01 at gmail.com> wrote: > Please can I formally propose that, for Linux: > 1. A Pragma or other compile-time option is created such that SQLITE_OK is > not issued on file writes/modifications/deletes until the hardware > indicates that all associated Directory syncs, etc., are done.
https://www.sqlite.org/src/timeline?y=ci&c=af92401826f5cf49e62c > 2. Since the absence of 1. appears to break the 'D' in ACID, the option is > set on by default, No. This would be a serious performance hit for billions and billions of devices that have soldered-in batteries and generally do not care about durability in the event of an OS crash. Remember that SQLite has never synced directories like this before in its 15+ year history and that was never a problem for anyone until a few weeks ago. If handset manufacturers, and/or OS vendors, want to make it the default on their systems, there is a compile-time option for that: https://www.sqlite.org/draft/compile.html#extra_durable -- D. Richard Hipp drh at sqlite.org
