Date: Fri, 20 Dec 1996 03:48:07 -0600 From: "Jacob Suter" <jsuter@intrastar.net> To: "Nate Williams" <nate@mt.sri.com> Cc: <isp@freebsd.org> Subject: cyclades data loss? (was: Re: attachment via pop using qpop problems) Message-ID: <199612200953.DAA24824@intrastar.net>
next in thread | raw e-mail | index | archive | help
> > We're using qpop 'popper' to deliver pop3 services for our customers. Some > > have recently had problems with both Netscape and IE3 mail not being able > > to download (timing out) on large file attachments. On our LAN, I have > > sent myself multi-megabyte files via mail and had NO problems recieving > > them with both Netscape 3.01 and IE3 for Windows95. > > Boy, we've had nothing but good luck with the same setup locally. The > clients are connecting over modems using Netscape and Eudora, but I've > never had any sorts of losses or timeouts that weren't due to crappy > lines, and qpop does a very good job of making sure email doesn't get > lost in that case. Well guys, I think I found the problem... After doing some stuff on my shell account (via 2400 from a friend's house) I think I can say I am losing data somewhere... I've found in the past that the Cyclades card apparently likes 57.6k a lot more than 115.2k (I don't think its the Cardinal 28.8ks, since I do not have this problem dialing through my portmaster, and all lines on it are at 115.2k using the same exact modems (even tried a swap tonight). I don't get many errors with ppp... probably because its only moving data at 576 bytes+headers without an ack... > I'm running qpop-2.2 on a 2.1.6.1 system w/sendmail 8.8.4, so it sounds > very similar to your setup. I've pondered upgrading just for the syn protection... :) > My boss even runs it over the internet when he's in Menlo Park, which is > *much* worse than any modem with timeouts & drops. Our lines here are consistantly bad, low fidelity, not noisy, so people connect at around 26.4k as a max, usually 24.0k and stay that way... I haven't seen any connection problems with the people in question (and I was getting no line noise on the 2400 that I could tell, just plain data loss). So, what can I do about this? I am using the Cardinal MVP288XF2 on these lines, with a Cyclades 8Yo card (the non-expandable w/ proper model signaling). I guess I could go down to say, 38.4k and make everyone slow down a bit :) anyone else seen this problem? TTYL JS
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199612200953.DAA24824>