Felipe W Damasio wrote:
> Hi Again,
>
>
> 2010/6/10 Felipe W Damasio <felipewd_at_gmail.com>:
>> Is this true? If I don't add "ignore_expect_100" the HTTP request
>> with it will _always_ fail?
>>
>> If this is right, why is that? I don't get it.
>
> I checked the RFC2616 that the ignore_expect_100 directive talked
> about, but there it contains:
>
> " - If a proxy receives a request that includes an Expect request-
> header field with the "100-continue" expectation, and the proxy
> either knows that the next-hop server complies with HTTP/1.1 or
> higher, or does not know the HTTP version of the next-hop
> server, it MUST forward the request, including the Expect header
> field."
>
> http://www.w3.org/Protocols/rfc2616/rfc2616-sec8.html
>
> So shouldn't squid (somehow) know that the www.megaupload.com
> webserver complies with HTTP/1.1 so that it can forward the request
> with the header? Even it doesn't (or can't) know, shouldn't it forward
> it as well (falling on the "does not know the HTTP version of the
> next-hop server" option)?
Squid knows itself to be a HTTP/1.0 hop which does not support the
100-continue responses.
The RFC section you quoted was relevant to known HTTP/1.1 hops.
*** Note the "1.1".
The next bit in RFC seems to be overlooked by most people and describes
the 417 error REQUIRED for known-HTTP/1.0 hops.
** Note the "1.0".
Besides which, I believe the invalid part is the software expecting a
100-continue to ever happen to a zero-length POST. There is NOTHING left
to continue sending.
Amos
-- Please be using Current Stable Squid 2.7.STABLE9 or 3.1.4Received on Thu Jun 10 2010 - 05:50:42 MDT
This archive was generated by hypermail 2.2.0 : Thu Jun 10 2010 - 12:00:03 MDT