I'm curious about the claim that case will outperform decode (well, not entirely in this case as decode's a function call - but it could be handled as a internal expression by the query compiler and I'm surprised it isn't, speaking as a grizzled compiler writer myself).
I'm even more curious about the claim that coalesce will outperform nvl. These are both function calls, after all.
So a pointer to documentation supporting this argument would be deeply appareciated, if it isn't too much trouble.
As far as query file choosing based on RDBMS version ... I've thought about it. Mostly I'd like to avoid it if possible. If the query's there for "old Oracle" syntax, which we need for 8i, I don't mind using it for 9i unless there's a compelling reason not to (performance, apparently, according to your information, but I'd like confirmation before we spend time worrying about it).
