I needed to write a script to monitor a site relied on by a production process where I work. Objectives were simply to measure load time of 3 pages, and send an email if it loaded longer than a variablized time threshold or returned a network error. My first thought was VU is perfect for the job since its measuring raw protocol data. However when I looked at the response times they were at least double the visual load time which I timed from initial address request til the globe stopped spinning (does anyone contend that network activity occurs after the globe stops spinning? good.) I set all delays and thinktime to 0. All connections were keep alive so there was no way to adjust the code to send concurrently(by putting all nrecv's after all the requests), thus having to send and recieve commands sequentially as the browser would have to do. I ended up doing it through a gui script which was probably easier anyway becasue of accessing oulook through ole2, but still the question remains: Are those kinds of differences showing up in actual performance tests? Somebody please restore my faith in VU(regardless, the Velvet Underground will still always have my utmost respect)