Date: Tue, 7 Feb 95 9:13:04 MST From: terry@cs.weber.edu (Terry Lambert) To: ponds!rivers@dg-rtp.dg.com (Thomas David Rivers) Cc: freebsd-hackers@freefall.cdrom.com Subject: Re: chat(8) improvements for SL/IP dialout. Message-ID: <9502071613.AA11843@cs.weber.edu> In-Reply-To: <199502041353.IAA19608@ponds.UUCP> from "Thomas David Rivers" at Feb 4, 95 08:53:10 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > Or you could do what everyone has done since time immemorial and open > > with a sleep the device so that the chat closer is not the final closer. > > > > Then the problem that the patch supposedly fixes will never occur. > > Yes, I thought about that, and considered it ugly... Well, it is ugly; mosly because the chat is doing something it shouldn't be doing and because the serial port is acting wierdly unlike those on say a Sun machine. A lot of this comes back to the partial open hack, which I guess people hate because it has the work "hack" in it, an the need for an open flag that completes the open when RI is triggered instead of DCD going high, which is what is needed for bidirectional and 'mgetty' type use, as well as for the secure callback scenario. You see, the connection between the open descriptor and the modem should not be severed as a result of DCD loss... the modem is not the controlling tty for the process. Currently, there are only two UNIX clones that implement connection loss (requiring reopening the port). The first is SCO Xenix 2.3.0 and the second is SCO Xenix 2.3.1. This is because I didn't report the problem until 2.3.1, so it wasn't fixed until 2.3.2. > My internet provider charges by the minute, thus, I'd like to connect > to the modem and *not* redial. (Hmmm... perhaps a flag on slattach > that says "this is how to dial the first time, but don't bother redialing > if Carrier Detect goes down".) > > So, I had planned on doing: > > chat -c "slattach ..." ... > > to start the connection, (setting the modem to disconnect after a minute > of no activity). Have sendmail do its thing, then when the modem disconnects, > use the slattach option to run the "shutdown-the-link" script. Thus, > since the shutdown-the-link script would have to go find the sleep and > kill it, I didn't like that idea. Well, the startup link could fork first thing, so it's possible to save the PID for, or even pass it to, the shutdown. I've often thought that what would help even more than this type of thing would be turning on digesting and building a digest exploder into an alias at the target machine. This would also have the effect of making it convenient to gate the lists to local news groups instead, as well as reducing the load on the mailing list machines (Andrew Tridgell did this for the SAMBA list). I guess the real answer is that BSD is still not well suited to intermittent connectivity (nomadic computing) or no connectivity (networking built in and huge kernel as a result by default). Terry Lambert terry@cs.weber.edu --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9502071613.AA11843>