Greetings. It appears that the problem may have been caused by a difference in MTU between the modem on my Linux box and the modem on the ISP's Linux box. MTU matched everywhere on my side of the link (at 1500), but not on the ISP's modem. So I changed everything on my side (ppp0, eth0 and my Win95 client) to 296, which is the same as the ISP's, and everything works fine. The web pages I could not access previously now fly past (well, as fast as you'd expect from a 33600 baud modem, but quite acceptable). I had a few suggestions about checking MTU, but I didn't think to go as far as I did until I noticed sendmail was reporting "collect: I/O error on connection" messages fairly consistently for several sites that were trying to send us email. A quick look at the sendmail FAQ web page (fortunately one of those I could access) said that this message might appear for a small number of sites, and may even not appear depending on the size of the emails being transferred. The FAQ suggested checking the MTU between the client and host systems. I did and they were different, and the rest is history. Recent history, granted, but history has to start somewhere. :) Next question. I have seen documented in several places that the MTU on the ppp0 and eth0 interfaces must be the same due to a bug or something in the Linux kernel (I have kernel 2.0.33). I have also seen a patch which is supposed to fix this, but the same site that had the patch noted that all patches had been incorporated into kernel 2.0.33. So the question is, given that I did seem to experience some problems, do I need to install the patch on my 2.0.33 kernel and would doing so help? Btw I checked my kernel configuration and IP_ALWAYS_DEFRAG *is* set. Many thanx to those who responded to my first question, especially Matt Roberts. Dave ---------- From: Dave[SMTP:[EMAIL PROTECTED]] Sent: Friday, 29 May 1998 8:48 To: 'Linux_masq' Subject: [masq] masqueraded http slows down Greetings. I have finally got my Linux box connected to our ISP and basically working doing masquerading amongst other things. Not through problems, just lack of time on my part. (Or on my hands.) Slackware Linux 3.4, kernel 2.0.33 (compiled with IP_MASQUERADING etc. etc.), um, pppd 2.2.0 etc. The rest is a standard Slackware 3.4 distribution. Running on a 486DX4/100 16M RAM, 38400 modem. During tests the CPU was idle about 97% (according to vmstat). The client is a Windows 95 pc with the Linux box configured as the default gateway and dns. Eventually about 8 PCs will be connected to the internet via the Linux box, internet usage is not huge tho. DNS works fine. pppd works fine. Email works fine. ftp works fine (provided I load the appropriate module :) http (using Netscape and Internet Explorer) works fine ... but ... after a while, usually around 15 or 20 minutes, loading web pages will slow to a crawl. At first response is fine, the modem lights flash and there is lots of receiving. But after this there is no activity. Occasionally the send light flashes. Minutes pass, then the receive light flashes a couple of times. According to Netscape transfer speed is something like 9 bytes/sec, that's just before the transfer stalls. At the same time tho, other things like ftp, traceroute, pings, emails still seem to work fine. MTU is 1500, ftp transfers of megabyte files happens very quickly and very smoothly. Any ideas about what might be going wrong and/or what I can do to fix it are more than welcome. tia Dave --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] For daily digest info, email [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] For daily digest info, email [EMAIL PROTECTED]
