Re: [squid-users] Frustrating "Invalid Request" Reply

 On Tue, 22 Feb 2011 17:24:39 +0200, Ümit Kablan wrote:
> Hi,
>
> 2011/2/21 Amos Jeffries wrote:
>> On Mon, 21 Feb 2011 16:19:53 +0200, Ümit 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=1ψ=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?

 Squid needs to be configured via the http_port to know what mode/type
 of traffic it is going to receive. The browsers need to be sending the
 right type as well.

 There are a number of workaround, So we are at the question of what
 exactly are you trying to do with the traffic? what does your goal look
 like?

 Amos
Received on Tue Feb 22 2011 - 21:04:33 MST