> Amos Jeffries schrieb: >> John Mok wrote: >>> Hi, >>> >>> I am using Squid3 nightly built 20071026 running on Ubuntu 6.06 LTS >>> with the compilation options :- >>> >>> ./configure --with-pthreads --enable-icap-client >>> >>> I tried both (i) the configurations with default option, or (ii) >>> icap-enabled options, the Windows client failed to get WindowsUpdate >>> (see the following log). >>> >> >> Where is this failure you speak of? >> The log you posted showed a proper link to WindowsUpdate, with all >> the static content coming from cache (TCP_HIT/TCP_IMS_HIT) and the >> dynamic pages and updates being brought in from M$ (TCP_MISS) >> >> If your client got the custom M$ "Windows Update failed" page. >> Then I suspect you have overlooked a M$ nasty: >> WU requires an HTTPS 'validation' test. >> You MUST permit an HTTP 'CONNECT' request to 65.55.184.125:443. >> (the IPA being that of www.update.microsoft.com from your current >> location) >> >> This bypass needs to be made on your firewall. WU will NOT always >> attempt it through the configured proxy :-( >> >> The best you can do is bypass it at the FW and also configure the >> proxy manually in IE, then run "proxycfg -u" in command line on the >> windows box, and hope that the particular box update level will use >> the proxy for it. >> >> >> Amos > Sorry Amos, > > the problem exists! It appeared with squid 3.0 RC1. After the downgrade > to 2.6 (urlgroup is missing in 3.0) I don't have any problem with WU. > The bypass in the firewall is not needed for proper operation. Lucky you, looks like you have a good up-to-date user base then :). Mine have trouble in WinXP SP1 and some earlier versions of the ActiveX WU'er they call MicrosoftUpdate. GenuineAdvantage my a*%#. > > Reinhard > IIRC others earlier found that WU used range requests to speed downloads. I have never confirmed this myself. Anyway, a bug has just been found in 3.0.RC1 that caused certain range requests to close prematurely. http://www.squid-cache.org/bugs/show_bug.cgi?id=2116 A fix has been incorporated in the next daily snapshot. If anyone is having this problem with 3.0.RC1, please give the 31 Oct or later snapshots a try and see if that fixes your problem. If its still present it will need to be reported as a bug with traces, etc. Thank you. Amos
