I presume you are referring to this sort of thing: http://grotto11.com/blog/slash.html?+1039831658 http://grotto11.com/blog/slash.html?+1039831658 It is perfec

I presume you are referring to this sort of thing: http://grotto11.com/blog/slash.html?+1039831658 http://grotto11.com/blog/slash.html?+1039831658 It is perfectly within spec to send data on a half-closed TCP connection, from the other end. That is exactly what the shutdown(SD_SEND) socket call is for. It signals the stack to send a FIN so that the other side can send data but knows to not expect more. They were probably implementing a version of HTTP 1.1 (before it became a formalized standard) that did not break existing clients (or servers) but speeded up theirs. BTW, I find it hard to believe IIS was sending more data after sending a FIN. Let's just say I worked in the vicinity of the Windows networking org in those days. I know it's not a popular history to believe among the SV crowd, but Microsoft in those days would bend over backwards for ISVs (Independent Software Vendors, or companies that wrote software for Windows) if they just whispered "your products are using undocumented APIs that I don't have access to". Regardless of whether that claim was true or not, they would work overtime to dispel the notion. The were just getting put under the anti-trust microscope, with a consent decree signed in 1994 and a bigger suit was obviously coming soon. Of course, when google does something similar with QUIC/SPDY, it's the second coming of Christ. Definitely not a way to boost their browser/core business. edit:typos