[e2e] RFI: Microsoft accused of TCP standards violation

[Marc Slemko]

On Mon, 6 Jan 2003, Jianping Pan wrote:

>
> It is interesting that we observed the similar behavior
> for Netscape Browsers with an Apache Web server. See the
> attached tcpdump and httpd.log for reference. It was
> done a bit long ago and I do not have their software
> version numbers handy now. Might be able to dig more.
>
> Initially we thought it might be a bug in Netscape. Now
> seems that it was *designed* to compete Microsoft's IE
> with Microsoft's IIS.

Nope, that is just a bug in Netscape where it doesn't
pay attention to the server closing the connection.

It can never gain any performance advantage out of
behaving in this manner.  It should always notice when
the server closes its half of the connection, and complete
the close.  The server isn't going to accept any more
requests on that half closed connection, and even if it
did then it couldn't send any responses.

See a document that I think I mostly wrote a long time ago
that talks about this:

	http://httpd.apache.org/docs/misc/fin_wait_2.html

(yes, it is an old document that hasn't been updated in a long time)