-> Linux. This seems to imply that it's a problem with the acroread plugin, -> and not with the OS or the Browser. Not necessarily. The Acroread plugin fires up as soon as it sees the Content-Type: header. The browser is still the one managing the TCP/IP and HTTP connections. But Acroread won't display anything on the screen until it has the entire document--this is different from HTML, where browsers try to render as much as they can as soon as they can. It might be a problem with HTTP 1.1 chunks... If you have the time, do this: Do an ethereal sniff of a single (failed) PDF request, save that as a parsed, human-readable text file (I don't remember the ethereal options), and then post it somewhere via HTTP and send the URL to the mailing list. I will download it and take a look at it. --Derek