On Thu, 9 Dec 2004, Garrett Rooney wrote:
But it's temporary allocation. You create a temp subpool at the beginning of the function, use it for the parsing and whatnot, then kill the pool at the end.
Isn't that significant extra overhead compared to just passing in an existing process->pool and request->pool?
I'm not sure that it would be measurable, especially considering that you're about to make a tcp connection to a relational database, something that will almost certainly take longer than allocating a new pool ;-)
Feel free to benchmark it though.
-garrett
