Date: Fri, 19 Apr 1996 01:52:09 -0700 From: David Muir Sharnoff <muir@idiom.com> To: joerg_wunsch@uriah.heep.sax.de (Joerg Wunsch) Cc: ache@astral.msk.su (=?KOI8-R?Q?=E1=CE=C4=D2=C5=CA_=FE=C5=D2=CE=CF=D7?=), nate@sri.MT.net, joerg_wunsch@uriah.heep.sax.de, scrappy@freefall.freebsd.org, freebsd-bugs@freefall.freebsd.org, core@freebsd.org Subject: Re: Re: bin/1019 Message-ID: <199604190852.BAA03814@idiom.com>
next in thread | raw e-mail | index | archive | help
* David, what do you think about reworking it? * * I think the summary of the opinions is: * * . no strong opposition against having this as an (optional) feature, * . it must be safe to use however, i.e. the chance of a misdetection * should be minimal. Currently I detect packets with two methods: 1. I look for the character 0x7e followed by 0xff. These are specified by the PPP protocol as the characters that start PPP packets. 2. I look for the following bit of junk " } } " that just happens to occur somewhere in the middle of PPP packets. Why did I take this approach? Because getty flushes buffers. In order to make a more deterministic getty, I would need to either: 1. Teach it more about what are valid PPP packets; or 2. Keep it from throwing away any input so that it can be sure that it's matching against the beginning of a packet. The way I figure it, even with my current methods, the chance of a noise-caused false positive is slight: First, this is in getty, not login so once any login name has been entered the detection will no longer operate. Second, most modems in use these days do error correction so line noise isn't as likely. Third, it's optional and defaults to off so it's only going to be on in situations where people are calling up with PPP. This means that the only people at risk of a false positive are those with old modems connecting to ISPs that are servicing a crowd that's using PPP. This group is small, but it isn't zero. Okay, what's the chance? I'm not sure. When I've used modems without error control, I've sometimes seen patters like " } } ", so I'm not sure that that one is impossible. Even though it's longer than the 0x7f 0xff pattern, I think it's more likely. As for me doing a better job of detecting it, I'm not sure. Certainly not immediatly. I think a more interesting thing would be to work on either mgetty or faxgetty and teach them to do it. I would like to be able to receive faxes too! -Dave --BAA09283.829903207/freefall.freebsd.org--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604190852.BAA03814>