Hi, 2011/2/21 Amos Jeffries <squid3@xxxxxxxxxxxxx>: > On Mon, 21 Feb 2011 16:19:53 +0200, Ãmit Kablan wrote: > >> >> and this works fine. The localnet counterpart: >> >> ------- >> GET >> >> >> /search?hl=tr&source=hp&biw=1276&bih=823&q=eee+ktu&aq=0&aqi=g10&aql=&oq=eee&fp=64d53dfd7a69225a&tch=3&ech=1Ï=6UBOTbHmCtah_Aa2haXRDw12969740590425&wrapid=tlif129697480915821&safe=active >> HTTP/1.1 > > Note the missing http://domain details in the URL. This is not a > browser->proxy HTTP request. It is a browsers->origin request. > > IIRC interception of this type of request does not work in Windows, since > the kernel NAT details are not available without proprietary third-party > network drivers. Look at WPAD configuration of the localnet browsers > instead, that way they will send browser->proxy requests nicely. Exactly! The working requests are all starting with http://domain/ as you mentioned. (I must say I couldn't capture loopback network packets in windows). I cant guess why firefox, ie, and chrome are sending protocol://domain'less requests when you hit enter but send correct url when google scans for autocompletion. I looked for advanced options but I couldnt get anything either. Have you got an idea of a workaround? Is it possible to tell Squid-conf to assume an exception for a Host (e.g. www.google.com) and if encounters a protocol://domain'less url: just concatenate two? > > Amos > Thanks for your attention, -- Ãmit
