IE4 "hangs", with TCP_MISS/500||504, HTTP POST, and IIS.

OK.

Here's the situation... I've read the FAQ, researched the mailing list
archives and DejaNews, et. al. We've got Squid 2.2-STABLE4 running on Linux
2.2.10 kernel. Users access all http, https, and ftp URLs through this
single Squid proxy. All users run IE 4 (which rev, I dont know, possibly
many revs). The problem:

Certain HTML documents which contain POST forms (for example, the
subscription form at
http://www.winntmag.com/OurProducts/Sub/Index.cfm?Action=Subscribe&Code=99IN
XTUP, or for another example the "send" button used for a new message under
MS Outlook Web Access) cause IE4 to "hang" (no error response from Squid)
when the user presses the submit button. In the access.log, an entry is
generated that looks something like:

936152282.992 901930 172.21.10.50 TCP_MISS/504 0 POST
http://www.winntmag.com/OurProducts/Sub/Index.cfm? - DIRECT/www.winntmag.com
-

OR

936152338.670 896169 172.21.10.50 TCP_MISS/500 0 POST
http://www.winntmag.com/OurProducts/Sub/Index.cfm? - DIRECT/www.winntmag.com
-

(note the difference is the return code - 500 vs 504, either is returned,
same symptoms from user perspective, however).

The problem does not happen when a user submits the form via Netscape 4.5.

A Netscape user's entry in access.log looks like:

936151112.151 700 172.21.10.50 TCP_MISS/302 450 POST
http://www.winntmag.com/OurProducts/Sub/Index.cfm? - DIRECT/www.winntmag.com
text/html

.. and the Netscape user is able to access the page generated by the POST
request.

The only other constant I've sort of been able to determine is that the
server is always running IIS in the problem cases. In the case a user of IE
submits a POSTed form to a server running, for example, Linux, the POST
always works.

Ugly. I wish it was even a remote option to suggest switching to Netscape,
but its really not.

The real kicker is that this problem did not surface under the combination
of Squid 2.0 and Linux kernel version 2.0X (we only recently upgraded the
proxy server to 2.2-STABLE4 and 2.2.10 from an older setup).

Additionally, and I don't know if it's meaningful, but in cache.log, we get
errors occuring every minute like this:

1999/08/31 23:28:06| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:29:06| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:30:07| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:31:07| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:32:07| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:33:08| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:34:08| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:35:08| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:36:09| clientReadRequest: FD 18 Invalid Request
1999/08/31 23:37:09| clientReadRequest: FD 18 Invalid Request

The filedescriptor number does change, but the error is constant, this was
just a log snippet.

Little help?
Received on Tue Aug 31 1999 - 21:59:48 MDT