Thanks for your responses.
I'm not using cursors or anything fancy. The expected behavior (as far as I can tell) for a .NET database driver is to read one row at a time from the database and make it available. There's even a standard API option for fetching data on a column-by-column basis: this allows the user to not hold the entire row in memory (imagine rows with megabyte-sized columns). This makes sense to me; Tom, doesn't the libpq behavior you describe of absorbing the result set as fast as possible mean that a lot of memory is wasted on the client side? I'd be interested in your take on this.
I can definitely appreciate the complexity of changing this behavior. I think that some usage cases (such as mine) would benefit from a timeout on the time until the first row is sent, this would allow to put an upper cap on stuff like query complexity, for example.
Shay
