Conference azur::mcc

A single AM can deal with any number of Global objects.  I am currently
working on a project that has a large AM that supports 4 global objects.

There is no restriction that I know of that would limit the number of different
objects you could place in the dispatch table, other than filling it up.

What defines the global entity is the *MSL*.

The AM code is (essentially) just a collection of entry point routines that
you've cobbled together into one binary.   These routines just get called to do
something, do it, and return the status.  They have no inherent knowledge, per
se, of what the entity structure(s) is/are.

/doug

	RE: .1 and .2,

	OK, so wha I would like to do is construct a block diagram of a network
element, and then place individual icons on the block diagram to simulate the
individual network element components.  Is this possible?

	Carl

RE: Dougs comment.

Doug, The AM's have embeded in them a HUGE amount of knowledge about what they
 support.  Only a few do not (e.g. are truely generic).  SNMP and Station
 are examples.

Phase 4, TSAM, SNA, Vitalink, Net Professor, DEC Bridge, and Phase 5 all have
 lots of hard coded assumptions about the entity(s) they control.

RE: .3

SNA (and what I am working on) currently do this by making selected
 child entities or network elements appear as global objects, too.
 Actually, in the SNA case, Front-Ends and PU 2s are viewed differently
 depending on if you are looking at the Mainframe's view of the element, or
 asking the element what it thinks it looks like.  That has led to more than
 a little confusion from time to time, until you get the hang of it.  Then it
 is all very reasonable.

So, I guess, in a nutshell, Yes, its fairly straightforward.  It has been done 
 a number of times already.

One nit deserves another.  ;^)

I said 'essentially' not 'exactly' nor 'only'.

/doug

Carl,

Putting the children of a global entity on the global entity map is 
possible and used by the SNA folks. 

The ramifications of this action is that you must explicitly register
the children.  The SNA folks using DNS V1.* found this to be a problem,
as they have thousands of child entities. There are implementation
problems with DNS record size restrictions and parents with lots of
children often blow out this boundary. 

As you are talking about TEMIP and probably are not using the namespace, 
this is probably less of a problem, but you still have to go through the 
hassle of registering each child.

You do have another option to just putting the children on a random map 
with a bunch of other global objects and the parent..  You can create a
domain to represent the parent.  This is like defining a container to
hold a bunch of objects. You then put the children as members of the
domain.  They still have to be registered, but the representation is 
more organized.

...kjn

hello, 

we have received through Stratacom a copy of some half-technical documentations
about a product called NetLabs/Manager.

It's a software running on top of almost every U*IX platform (unfortunately or
... hopefully not ULTRIX) and sold by a small californian startup company 
(NetLabs inc.).

Their package contains what they call "NetLabs/Vision Development Environment" :

"it provides a 4GL for the effective development of graphical network management
applications. The environment includes the NetLabs/Vision Device specification
Language (DSL), Compiler, and runtime engine. It creates Motif or Openlook UNIX
applications to manage SNMP devices on TCP/IP networks ....

... it provides an easy-to-use object oriented Device Specification Language 
(DSL) for rapid development of device management applications. A script written
in DSL contains the device description, device reporting formats and expressions
, and screen window layouts.
To create an application for a specific device, you compile your 4GL DSL script
and device MIB. The NetLabs/Vision compiler creates a device specific file 
(DSF). The DSF is independent of GUI, network management platform and hardware 
architecture, saving your porting costs.

When the user runs the application, the Runtime Engine takes as input the DSF
and a device image file. It gathers information from the monitored device using
its SNMP protocol modules and presents a graphical representation of the device.
..... "

I wonder whether it would be possible one day to define "graphical" attributes
in a MSL file such as the IMPM would use these attributes to display diagram
boxes of entities and children rather than just icons and text strings.

I know that nothing like this exist in the current protocols (neither CMIP nor 
SNMP V2), but this could be added after compiling a MIB for example.

This feature is really missing in our current product, isn't it ?

If you want to get a fax copy of the NetLabs documentation, please send me 
a mail with your fax number.

Christophe

	Thanks for the replies, at least I have some idea how it is done.

	Carl