Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!usc!nic.csu.net!beach.csulb.edu!csus.edu!decwrl!sdd.hp.com!zaphod.mps.ohio-state.edu!
 magnus.ircc.ohio-state.edu!csn!hellgate.utah.edu!fcom.cc.utah.edu!npd.novell.com!newsun!donp
From: donp@na.excelan.com (don provan)
Newsgroups: comp.dcom.lans
Subject: Re: Packet Drivers or NDIS Drivers for Netware
Summary: more bickering about packet drivers vs. ODI
Keywords: NDIS,Netware,Drivers
Message-ID: <1991Mar6.201431.18839@novell.com>
Date: 7 Mar 91 06:47:56 GMT
Sender: news@novell.com (
Lines: 43

The News Manager)
Nntp-Posting-Host: na
Reply-To: donp@novell.com (don provan)
Organization: Novell, Inc., San Jose, California
References: <1991Feb22.221957.18674@novell.com> <13782@asylum.SF.CA.US>
Date: Wed, 6 Mar 1991 20:14:31 GMT

In article <13782@asylum.SF.CA.US> romkey@asylum.SF.CA.US (John Romkey) writes:
>From what I understand, ODI doesn't provide any logical (IP, for instance) to
>physical (MAC) address translation. If it doesn't, then any protocol stacks
>running above it still have to do this and know about the physical media they
>reside on. In that case, ODI's only real difference is that it builds media
>headers. Without the address translation, I don't think you can really claim
>that it's media-independent, as some protocols like IP will still have to
>provide media dependent code in them.

The existence of ARP shows that IP itself isn't media-independent, so
of course complete media-independence for an IP implementation is not
possible.  Other protocols, however, such as CLNP, XNS, and IPX,
require no media specific address mapping because they pass media
addresses around at the network layer.  Here, ODI has an overwhelming
advantage because these protocols can, in fact, be completely
media-independent.

In other words, IP's use of logical addresses is not typical.  Even if
it were, there's a significant difference between identification of
the medium and knowledge of the medium.  ARP itself is interesting for
a few reasons:

1. Address mapping really *is* the only media specific information
required.

2. Address mapping is a rare event, so the media specific code isn't
invoked for each packet.

3. ARP is common across several media.  IP with ARP really *is*
basically media-independent for a wide range of media.  (I'm assuming,
of course, that ARP implementations tend to ignore ARP's hardware type
field, and i can't imagine why they wouldn't.)

Anyway, Keith's main point still stands: ODI and packet drivers are
complementary solutions to different problems, not competing
interfaces at odds with each other.
						don provan
						donp@novell.com