[rescue] AT&T 3b1 Starlan software

[Sheldon T. Hall]

"Jonathan C. Patschke" <jp at celestrion.net> writes ...
To: "The Rescue List" <rescue at sunhelp.org>
Sent: Thursday, February 13, 2003 3:32 PM
Subject: Re: [rescue] AT&T 3b1 Starlan software


> On Thu, 13 Feb 2003, Scott Newell wrote:
>
> > OK, this is what I wanted to know.  If software assumes that the floppy
is
> > always slot 6 and you move it, you're hosed, correct?  That's the big
> > drawback that I can see.  (And I realize it's only a problem for poorly
> > written code.)  Did they make allowances for cards that require more
than
> > one interrupt line?
>
> Why would a card need more than one interrupt line?  The card itself
> should only need one line to the CPU to say "Hey!  I have data!".  The
> notion of needing, for example, one IRQ per serial port is wasteful.
> It's not hard at all to multiplex interrupts for a multiport card.  The
> CPU can only service one interrupt at a time, so you really don't gain
> anything by having the serparate IRQs.

That worked pretty well, actually. I wrote a package for NBC that exploited
that feature; it would handle 10 serial devices on a '296 under MS-DOS.
Actually, it would handle many more devices, but I don't think any of the
250 or so PCs in the field used more than 10.

The hardware was the normal 2 PC serial ports (one IRQ each) and an 8-port
card that shared a single IRQ among its 8 ports.

Given that an ISA-based PC is really short on IRQs to start with, it seems a
bit thoughtless for every serial port to want its own.

Sorta like clues, I guess.  Not enough to go around, so not everybody gets
one.

-Shel