RE: NIM BoF

["Andrea Westerinen" <andreaw@cisco.com>]

Tom, I agree that TINA - more specifically its NRIM (Network Resource
Information Model) - is appropriate to consider.  However, TINA describes an
entire architecture and its implementation guidelines.  As such, its scope
is beyond NIM.  (IMHO, NIM is an OO model of a network and the
description/rendering of the model).  Also, since TINA is a consortium (a
.com), I did not specifically list it in my quick overview.  I only listed
the .orgs.  If folks are interested, you can go to www.tinac.com for more
info.

On the other subject, below - "protocols" - I think that I disagree, but am
not sure how wide or narrow your definition of "protocol" is.  Here is my
interpretation ... Protocols are about how data is sent, replied to, routed,
etc.  It is about very specific data in a very specific context.  For me,
NIM is (solely) about the data, starting with general concepts and
abstractions.  It is about the semantics of the data - describing the nouns,
verbs and relationships in a network.  This description should be generic
and not context/protocol specific.  However, it can and should abstract the
concepts in protocols and implementations.

Since similar concepts oftentimes repeat themselves in individual protocols
and implementations, wouldn't it be nice to know that "item A in MIB XYZ" is
the same conceptual entity as "item 23 in Protocol 456"?  Or, making it
harder, that these two pieces of data are related via a MemberOfCollection
association?  This is the distinction between an info model and a data
model. (The data model is implementation specific and likely to be a subset
of the info model.) In as much as a protocol defines and describes the data
that it is conveying (for example, DHCP), protocols are data models.  They
are a specialization and implementation of the general concepts in an
information model.

NIM is not about protocols but concepts, and the right abstractions and
specializations.  For example, think about finding all synonyms for fan
(fan, cooling device, temperature control) in a MIB, and all synonyms of
"good" (good, nominal, functional, working, OK, satisfactory) to figure out
if you have a cooling problem in a piece of hardware.  This should be easy,
but it is not.  If you look at protocols in isolation, you may not even see
a fan.  If you look at a customer trying to manage his/her network, you can
bet that they care more about a failed fan than about other data - if it
means that a high traffic router is now shutting down, due to temperature
issues.

An information model provides the semantic backing behind multiple protocols
and implementations - and allows for consistency across them. Also, it
organizes data, allowing the "bigger picture" to be created. IE, the
customer's network consists of lots of hosts, clients and network devices
with lots of QoS, IPsec, miscellaneous MIB, DHCP and other data.  You can
organize and correlate all this info, if the underlying semantics are
well-understood.  Or, you can spend lots of money trying, if the semantics
are each independently defined.

Andrea

-----Original Message-----
From: Tom Scott [mailto:vedatel@hotmail.com]
Sent: Wednesday, August 02, 2000 10:01 AM
To: andreaw@cisco.com
Cc: nim@psg.com
Subject: NIM BoF


Andrea,

I agree with the comment at this morning's NIM BoF session in regard to the
need to explicate the NIM concepts and model within the general context of
protocol development. To be more specific about what is meant by "protocol",
I refer you to the following two works:

Elements of Network Protocol Design, by Mohamed G. Gouda (Wiley, 1998)
http://www.amazon.com/exec/obidos/ASIN/0471197440/o/qid=965232112/sr=2-2/102
-9323651-8129762

Object-Oriented Network Protocols, by Stefan Boecking (Addison-Wesley, 2000)
http://www.amazon.com/exec/obidos/ASIN/0201177897/qid%3D965232523/102-932365
1-8129762

I also suggest that you list the modeling work of TINA (Telecommunications
Information Networking Architecture) and ODP in your slides for the sake of
completeness. There are five models in TINA/ODP that the group should be
aware of: enterprise/business, informational, computational, engineering,
technology. It may be valuable to compare/contrast these five models to the
two types of models discussed this morning: information and data. (I'm going
from memory here, so perhaps someone from TINA/ODP can correct me if I've
named the five models incorrectly.)

--Tom
---------------------------------------------------
Tom Nelson Scott             Vedatel Co
1411 Sheffield Dr.           Bowling Green OH 43402
"In IP We Trust"   "E Pluribus Unix"   "Java Rules"
---------------------------------------------------


________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com