[squid-users] Load-balanced Squid, "always_direct", and Cached Content

If "always_direct" is used for systems in a specific domain, will a Squid
proxy query its peers for cached content?

   (1) WPAD is defined in the local DNS zone to support the automatic
        detection of web proxies. Assuming that automatic detection is
        enabled, Linux- and older Windows- based will locate and use the
        proxy.pac/wpad.dat configuration file; however, WindowsXP-based
        systems will not.
   (2) DHCP is configured to pass Option 252 to pass the URL for the web
        proxy automatic configuration script. Assuming that automatic
        detection is enabled, WindowsXP-based systems will only use the
        proxy.pac/wpad.dat file when configured for DHCP and Option 252
        is defined to the URL.
   (3) The same proxy.pac/wpad.dat file that defines a simple hashing
        algorithm to balance the load between two Squid proxy servers.

When a Mozilla-based browser (FireFox, Mozilla, Netscape, etc.) is used,
load-balancing works as expected. When Internet Explorer is used, the
load-balancing fails. If a specific web server has never been accessed,
the hashing alforithm is used to select the web proxy. However, once a
proxy selection has been made, Internet Explorer will use that proxy for
all accesses to that specific web server.

The load-balanced proxy servers have a common parent that sits on the
network security perimeter to provide access to Internet content. When
content is being retrieved from the Internet it doesn't matter whether
one Windows system selected proxy A and another selected proxy B. If the
content is cacheable, it will be retrieved from the parent's cache.

The parent proxy can access both external Internet content and internal
corporate content over our wide area network. At present, "always_direct"
is only used on the load-balanced proxy serversto access web servers in
our local facility. I would like to use "always_direct" on the proxies
for all internal corporate web content.

If all browsers used the hashing algorithm for all requests, there would
be no question about using "always_direct" as the browser would select the
proxy that would have the content cached. The question is does the proxy
query its peers before applying the "always_direct" rule? If not, does
this behaviour change when using NetDB?

Merton Campbell Crockett

-- 
BEGIN:				vcard
VERSION:			3.0
FN:				Merton Campbell Crockett
ORG:				General Dynamics Advanced Information Systems;
				Intelligence and Exploitation Systems
N:				Crockett;Merton;Campbell
EMAIL;TYPE=internet:		mcc@CATO.GD-AIS.COM
TEL;TYPE=work,voice,msg,pref:	+1(805)497-5045
TEL;TYPE=work,fax:		+1(805)497-5050
TEL;TYPE=cell,voice,msg:	+1(805)377-6762
END:				vcard