On Sat, 22 Jan 2000, Henrik Nordstrom wrote:
> > It would be much more elegant to force a close of the persistent
> > connection in this case.
>
> Squid must eat the message data, or the client will receive a TCP reset
> and will never see the 401 reply sent by Squid.
>
> If the connection is closed or not does not matter much, other than that
> some clients are broken when sending POST/PUTs.
Eating the message body is fine. I was more referring to the ugly part
of scanning for newlines and looking for words that seem to be
request methods.
Can't we send connection: close on the "auth required" reply? There
should be a subsequent request already on the socket only when the
client is pipelining.
Doesn't 8.1.2.2 of RFC2616 come into play in the case where
Squid resets a connection like this? The client should
retry its request.
Received on Mon Jan 24 2000 - 08:05:16 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:20 MST