Re: [sqlite] rollback/replay journals and durability of the most recent transaction

2008-07-02 Thread Karl Tomlinson
Thank you for your comments. Karl Tomlinson wrote: One thing I'm trying to understand is whether there was a reason for SQLite choosing to use a rollback journal (of the steps to undo a transaction) rather than a replay journal (of the steps to perform a transaction). I didn't make this

Re: [sqlite] rollback/replay journals and durability of the most recent transaction

2008-07-02 Thread Karl Tomlinson
Igor Tandetnik writes: Karl Tomlinson wrote: I really meant, as a first possibility, that writing the pages to the database itself would be performed during the commit process after syncing the replay journal (but the database need not be synced until the journal is about to be removed).

[sqlite] rollback/replay journals and durability of the most recent transaction

2008-07-01 Thread Karl Tomlinson
I've done a little looking into journals, fsyncs, and filesystems recently. One thing I'm trying to understand is whether there was a reason for SQLite choosing to use a rollback journal (of the steps to undo a transaction) rather than a replay journal (of the steps to perform a transaction). In