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 3.1.0.16Received on Sat Feb 20 2010 - 10:10:09 MST
This archive was generated by hypermail 2.2.0 : Sat Feb 20 2010 - 12:00:05 MST