Subject: Re: A new user's comments, part 4
To: Scott L. Burson <firstname.lastname@example.org>
From: Terry Moore <email@example.com>
Date: 06/05/1995 11:44:10
> There are some difference in the tty driver vis-a-vis SunOS that are causing
> some minor problems. Programs like `tip' can't open the serial port if CLOCAL
> is not set, but, seems to me, it *shouldn't* be set on modem lines. I tried
> making `tip' open the port with O_NONBLOCK, but that is not sufficient -- it
> can open the port, but then can't write to it. I guess it will have to open
> the port with O_NONBLOCK, immediately set CLOCAL, then clear CLOCAL once
> dialing is complete (but only if it really did dial, and didn't just connect
> to the port). Really, I think I prefer the SunOS interpretation of CLOCAL,
> which appears to be just to generate SIGHUP on an on-to-off transition of DCD
> and to ignore DCD otherwise. But NetBSD's behavior is at least consistent
> with its documentation.
Probably only a minor thing: but I have a version of getty (uugetty) that uses
the lock files and allows things like kermit to work. Tip is, I agree, broken
with respect to modems; but using uugetty, I've sucessfully multiplexed kermit,
dial ins, & ppp dial outs without interference, and without source mods to either
the kernel or the utilities.
Some people have recommended switching to a ttn / cun scheme (a-la sunos), but
that seemed to still require lots of other things to get modified (tables,
defaults, and so forth); just writing a uugetty seemed to be much easier.
Patches available on request.