Subject: Re: serial line ideas
To: None <current-users@NetBSD.ORG>
From: Michael Richardson <>
List: current-users
Date: 12/27/1996 16:58:50

>>>>> "Mike" == Mike Long <> writes:
    Mike> USB is more like Apple's ADB than a RS232 serial link.  If
    Mike> or when it is supported, it will probably be only as another
    Mike> bus to which devices may be attached in the kernel.  USB
    Mike> interfaces will bear no similarity to a 8250/16[45]50 UART.

  I wasn't picking on the chipset.
  If you connect a USB modem, ISDN adaptor, etc... to the system, is
it going to show up as /dev/ttyXX, and that is *it*? Will the USB have
the notion of "control" channels as opposed to "data" channels?
  How will you toggle "DTR" to hang up the modem? What will that mean?
What about crtscts? Will that having meaning? I suspect neither will
mean anything, and we'll wind up having some kind of daemon that talks
USB protocol, and uses pty's to provide tty service.
  What *I'm* suggesting is that we should do the same thing for 16550s
connected to AT/Hayes-compatible modems. We should talk
*AT-command-set* protocol to the modem, and provide "tty" like access
via pty's (or some other kernel-copyin/out friendly replacement. It
could locate its ring buffers in user space via mmap() with semaphores
to protect it counters). 

  What am I trying to say here: is there interest in this? Are there
people interesting in contributing ideas, reviewing design? 
    Mike> I agree that trying to connect ISDN via an RS232 link is
    Mike> Wrong; but you should blame the hardware designer, not the
    Mike> operating system.  -- Mike Long <>

  Along these lines: are there people familliar with Intel's I2O spec?
Anyone doing any work with it? It suggests a higher level of I/O
device/OS compatibility on ISA/PCI buses to me.

   :!mcr!:            |  Network security consulting and 
   Michael Richardson |      contract programming
 WWW: PGP key available.

Version: 2.6.3ia
Charset: latin1
Comment: Processed by Mailcrypt 3.4, an Emacs/PGP interface