Received: (from majordom@localhost) by hyperreal.org (8.8.5/8.8.5) id RAA06789; Sun, 31 Aug 1997 17:19:57 -0700 (PDT) Received: from aaue.matrice.it (root@[194.185.177.194]) by hyperreal.org (8.8.5/8.8.5) with SMTP id RAA06778 for ; Sun, 31 Aug 1997 17:19:53 -0700 (PDT) Received: from localhost by aaue.matrice.it with smtp id m0x5KQA-000IS3C (Debian /\oo/\ Smail3.1.29.1 #29.37); Mon, 1 Sep 97 02:32 MET DST Date: Mon, 1 Sep 1997 02:32:46 +0200 (MET DST) From: Giovanni Maruzzelli To: "Philip A. Prindeville" cc: new-httpd@apache.org Subject: Re: [BUG] 1.3a2 on NT (no authorization? not closing socket ?) In-Reply-To: <199708312214.PAA27575@enteka.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: new-httpd-owner@apache.org Precedence: bulk Reply-To: new-httpd@apache.org On Sun, 31 Aug 1997, Philip A. Prindeville wrote: > > Giovanni Maruzzelli wrote: > > * when connecting from a Unix machine to an NT machine running > > apache-1.3a2 the sockets remains in a TIME_WAIT state. This > > not happen with windoz clients. > > Do you mean the sockets at the Unix end or the Windows end? > > Cheers, > > Ben. If I remember well, only at the NT side. > > Actually, a socket is *supposed* to stay in the TIME_WAIT state for 4 > minutes, 15 seconds. This is referred to as "quiet time". Since the > IP TTL is an 8 bit counter, that is supposed to be decremented by one > for each hop it takes *or* by the number of seconds the packet is > held in a router (min 1), then a looping duplicate or delayed out-of- > order packet can last up to 255 seconds (4:15) in the network, and you > need to cache the TCP Control Block in the closed state to make sure > that the same port doesn't get reused in that period and mistakenly > accept data destined for the previous instance of that socket. > > Make sense? > > But if the socket is staying in the TIME_WAIT state longer that 4:15, > I would say one end is doing a broken close (or getting disconnected > abruptly). > > -Philip > Don't take my poor english the wrong way, but not it don't make sense to me. Or, maybe yes, but there's something still to be explained to me: * in the unix version of apache I don't see socket thet remain in CLOSE_WAIT * the NT version behave as you say, but after a while it fill up all the available sockets and almost freeze (waiting the sockets close) * when you connect from a win client to an NT apache it don't live sockets in CLOSE_WAIT state * if you modify apache NT to setsockopt linger on lingering_timeout to zero and not to use lingering_close() it close the socket (but it is seen as a network error from the clients also if you can see the contents delivered via a perl homebrew client) For the auth problem, you can set auth on a location, but not for a directory. I'll check if the problem comes from the upper-lowercase stuff. ciao, -giovani *********************************************** * * * * * MATRICE srl * * Information Technology Solutions * * * * Giovanni Maruzzelli * * * * * * tel. (+39)(2)2952 0481 * * fax (+39)(2)2952 0528 * * maruzz@matrice.it * * http://www.matrice.it * ***********************************************