>The actual data transfer is not any slower, but Netscape seems to
>be reporting it as such. I'm not sure what Netscape bases transfer
>rate calculations on. I'm assuming a value in the Content-Length.
Weird. One thing I did notice, since Dean's buffering improvements,
is that a slow CGI script won't send *anything* until it either
completes or fills an 8KB (IOBUFSIZE) in send_fd. This is the one
area that I wouldn't mind a bflush between the header fields and the
body, but I don't run a site that depends on fast CGI buffering.
Hmmm, I wonder if your Content-Length just happens to be at the
256-257 bytes?
.....Roy
|