Re: NTLM changes in IE

Has this been revisited at all?

The suggested registry change seems to have made the POST problem go from
sporadic to slightly more sporadic, but it still seems to exist.

Thanks,
Jerry

----- Original Message -----
From: "Henrik Nordstrom" <hno@marasystems.com>
To: "Robert Collins" <robertc@squid-cache.org>; "Squid Dev"
<squid-dev@squid-cache.org>
Sent: Wednesday, November 27, 2002 7:04 PM
Subject: Re: NTLM changes in IE

> On Thursday 28 November 2002 00.44, Robert Collins wrote:
>
> > Its meant to be closed immediately - IE is not meant to send the
> > request body once auth is required (according to some MS article I
> > ran into).
>
> Well.. to play well with TCP/IP we MUST keep on reading the connection
> until the full request has been received or the connection is closed
> by the browser, or else the browser may receive a RST from us
> completely screwing up things to the point of even not allowing the
> browser to receive the 407 reply (timing dependent).
>
> This is true for all the other authentication schemes and/or methods
> (in fact any response generated by Squid), and SHOULD be true when
> using NTLM as well.
>
> The browser MAY abort the transmission of the body upon receiving the
> 407 reply or another error, but we cannot force the browser to do so.
>
> This is also one of the reasons to why "100 Continue" is added to the
> HTTP/1.1 specs, but the same problem still applies to some extent
> even when "100 Continue" is supported.
>
> Similar issues can be seen if the browser initiates request
> pipelining.
>
> Regards
> Henrik
Received on Thu Jan 02 2003 - 06:13:40 MST