No. When the _reply_ starts with "100 Continue" the remainder of the POST
may be discarded.
Reply ?! You are saying about answer which receives Squid from http server ?
so Squid than blocks the expect replays ? Right ?
----- Original Message -----
From: "Amos Jeffries" <squid3@xxxxxxxxxxxxx>
To: <squid-users@xxxxxxxxxxxxxxx>
Sent: Saturday, February 20, 2010 11:09 AM
Subject: Re: HTTP/1.0 417 Expectation failed
Riccardo Castellani wrote:
When receiving 417 it should be re-trying the POST without the expect
header, or the delays Expect adds to the system.
a- What means "or the delays Expect adds to the system" ? To wait answer
to Expect request ?!
Yes. There is a long (unknown length) delay waiting for the 1xx response
before the second half of the POST is sent. When there is no 1xx response
existing the delay may be very long.
The 417 shortens that by saying explicitly there will be no 100 happening.
I assume of course that the agent doing the POST can handle the 1xx
response wait properly. It seems not to be capable of the closely related
417 part of the requirements.
Squid other than 2.7 will not produce this error. They will just die in
unexpected ways if the 100-continue your software asked for actually
comes back.
b- You are saying Squid process (a version different from 2.7) will die
if it receives a request which contains "Exepct: 100-continue" ?
No. When the _reply_ starts with "100 Continue" the remainder of the POST
may be discarded.
c- But the 417 error is given from Squid or from remote Http server ? I
suppose, in this case, the error is given from Squid 2.7
Doesn't matter. It can be produced by anything along the route.
Squid-2.7 sends it because 2.7 is smart enough to know that _it_ can't
handle the 100 reply and what the reponse is supposed to be when that
ability is missing from any part of the relay.
Amos
--
Please be using
Current Stable Squid 2.7.STABLE8 or 3.0.STABLE24
Current Beta Squid 3.1.0.16
