If you have any influence over the site you're fetching data from, it would be a lot faster to write a CGI on their server to do the counting for you.

That said, if all you're doing is counting lines that contain "flist", you might as well do a read() on the socket to get a larger buffer, then grep the buffer for the "flist". Matching in huge files demonstrates a fast technique for grepping through a file without having to pull it all into memory.