externals and close_wait tcp connections

[kmr...@rockwellcollins.com]

Using both the svn 1.6.5 command line on windows and TortoiseSVN 1.6.5, I'm seeing
a TCP socket get into a CLOSE_WAIT state on the client and FIN_WAIT2 on the server
for each and every svn:external fetched during an "svn update".

Using process explorer, I can see a new TCP connection created for each
external fetched.

I get hundreds of these on the client machine and things just grind to a halt on the client
(Windows XP on a quad core machine) after enough of them are left in that state.

The server is Windows 2003 server running apache 2.2.13 and svn 1.6.5.  The
client is Windows XP SP3 running svn 1.6.5.

It appears this is due to the client?? not finishing the connection close.
(I may have misinterpreted some google results...)

Anything I can tweak (on the server or client) to help an svn update on
a working copy with thousands of externals to complete?

I have only tried neon, would serf be better/worse?

Could a misbehaving firewall between the two machines cause this problem?

Thanks!
Kevin R.

[kmr...@rockwellcollins.com]

Not a firewall issue, since I took that out of the equation.  I am
able to "fix" the behavior by setting "KeepAlive Off" in the httpd.conf
file, but that breaks sspi and kerberos authentication.

Kevin R.

[kmr...@rockwellcollins.com]

It appears similar to this:

http://subversion.tigris.org/issues/show_bug.cgi?id=1448

Can it really be an issue from 2003????

Kevin R.