There are some controversy about definitions of isolation levels in SQL standard. For example absence of mentioned phenomenas doesn't guarantees serialisability

There are some controversy about definitions of isolation levels in SQL standard. For example absence of mentioned phenomenas doesn't guarantees serialisability, as it shown in Berenson paper [1]. Implementation independent description of weak isolation levels can be found in Adya phd thesis [2] (and that's a tough read). Also nice evaluation of modern databases in terms of [2] can be found in Martin Kleppmann's github [3]. [1] https://www.microsoft.com/en-us/research/publication/a-critique-of-ansi-sql-isolation-levels/ https://www.microsoft.com/en-us/research/publication/a-criti... [2] http://pmg.csail.mit.edu/papers/adya-phd.pdf http://pmg.csail.mit.edu/papers/adya-phd.pdf [3] https://github.com/ept/hermitage https://github.com/ept/hermitage