Thanks, I got the part about the spec being ambiguous, am more interested in the "why" aspect, since the current behaviour seems intuitive to the name "repeatable read". But on closer inspection, I see PostgreSQL's repeatable read blocks phantom reads even though the ANSI spec permits that! I don't get why phantom reads would be acceptable under "repeatable read"... I probably should give those papers a read some time. But in the meantime, given the choice of phantom reads or G2-item, I think I'd pick blocking phantom reads. (It might be nice to have the option to choose though!)

In PostgreSQL's case, if they somehow made repeatable read to prevent G2-item without sacrificing the phantom reads, would that mean repeatable read is then "serializable" according to the ANSI definition?