Subject: Re: /dev/cuaXX (Was: Formal getty replacement yet?)
To: None <perry@imsi.com>
From: Greg A. Woods <woods@kuma.web.net>
List: current-users
Date: 12/20/1994 11:59:54
[ On Mon, December 19, 1994 at 21:35:06 (-0500), Perry E. Metzger wrote: ]
> Subject: Re: /dev/cuaXX (Was: Formal getty replacement yet?) 
>
> Because one might want to have a getty running on the line for when
> someone dials in, but also run UUCP outgoing on the same line. You run
> the getty on the device that only opens when it gets carrier, and it
> will pick up if anyone calls. You run the UUCP on device that opens
> when it has no carrier and its happy, too. The two can never step on
> each other because only one will open at a time. Its a marvelous
> hack. I've used it for years on Suns and I sorely miss it. Lock files
> for ttys are orthogonal and not a replacement.

But multiple /dev/tty's for for the same port is an icky kludge, and it
hides valuable information in the kernel.  (I.e. it's non-unix-like ;-)

User-level race conditions are avoided, but possibly at the expense of
kernel level race conditions being more possible.  The user-level race
conditions, and the complexity of using lock files, can all be
eliminated by writing a simple library routine ala SysV's dial(3c).

-- 
						Greg A. Woods

+1 416 443-1734			VE3TCP		robohack!woods
Planix, Inc. <woods@planix.com>; UniForum Canada <woods@uniforum.ca>