Subject: Re: Another changer, another changer problem
To: None <current-users@netbsd.org>
From: Greg A. Woods <woods@most.weird.com>
List: current-users
Date: 10/03/1998 21:17:10
[ On Sat, October 3, 1998 at 16:11:20 (-0700), John Nemeth wrote: ]
> Subject: Re: Another changer, another changer problem
>
>      In most aspects, I would consider IRIX to be fairly sane.  The
> main thing that gets me is its major lack of security out of the box.

As an aside I think the reason IRIX is insecure, even long after it's
been out of the box, is because it seems no IRIX systems programmer ever
heard of the idea of defensive programming, and they seem to like
setting the setuid bit as a quick hack to avoid doing things right in
the first place, and they've never heard of, let alone read, any guide
to writing setuid programs.  They should only be allowed to write in
Turing or some totally safe language.

>      The only way that should be used is physical slot.  Everything
> else leads to unnecessary complexity and confusion.

Yes, this is a good idea, but it's not necessary or possible on some
modern buses (though unfortunately on some where it's not possible it's
still necessary).

> } inspection, and by the kernel.  If some patch to your OS causes the
> } kernel to count from the other direction suddenly then that's something
> } you do need to be aware of, but it's not generally a problem.
> 
>      No, but it blows your whole argument.  If you can't depend on the
> things to stay put, then why bother using the location as the
> identifier?

Does not!

If you know which way your kernel counts then all you need to do is
count.  If the kernel suddenly starts counting from the other side then
you simply count from the other side too.  No magic necessary --
one-to-one correspondence remains.

>      As I showed above, the only reasonable and reliable way to handle
> cards is by the physical slot that it is in.  Having switches of some
> sort only leads to limitations of how many items of a given kind you
> can put in the machine and various other conflicts.

Hmmm... I've never seen much of a problem with using address switches
and such on VME systems.  Yes VME has timing problems that make it
difficult to put too many of a certain type of card in, or various types
of cards too close or too far from the CPU, etc.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>