Ever contacted the owners of a network stack driver, like Microsoft, Intel
and/or Broadcom ?
They should know that they are buggy and not ASF, special Microsoft should
know that some is buggy for HTTPD (only).
I still think:
Hardly believe that it is a driver problem, I guess more it is an http
design issue, seen it nowhere else for all that years.
I suggest to consult/hire Microsoft (believe is a ASF sponsor), they certify
the drivers.
Steffen
-----Original Message-----
From: William A. Rowe Jr.
Sent: Wednesday, February 15, 2012 7:00 PM
To: dev@httpd.apache.org
Cc: Steffen
Subject: Re: [VOTE] Release Apache httpd 2.4.1
On 2/15/2012 11:39 AM, Steffen wrote:
Blaming (stack) drivers with set to data:
Bug shows when tested with server rated cards from Broadcom and Intel with
old and new certified drivers on clean 2008 R2 and SP1 and XP and with
consumer rated cards like Realtek, and that on quite some different boxes.
So feels bug is not there, but you are the expert.
To be clear, again... Winsock 2.2 provides several mandatory features
which are integrated into the driver code;
1. recycling sockets (through TransmitFile and DisconnectEx)
2. accelerated kernel-based file transmission (through TransmitFile)
3. async accept-with-data (through AcceptEx)
When any one of these were incorrectly implemented by a network stack
driver, DisableWin32AcceptEx (and now AcceptFilter none) was necessary
to work around the broken driver. That feature has always been a gross
hack around what should "just work".
