WAL should supersede all these concerns about fsync. At the very least, it changes the tradeoffs enough that there's no point in designing performance improvements based on the old code... regards, tom lane
pgsql-hackers by date:
Соглашаюсь с условиями обработки персональных данных