Re: Using X.500 to determine presentationAddresses

[[e-mail]]

Peter et al,

My apologies for such a late reply. 

The open issue I see here is the naming of the machines. 
During the last month I have spent many hours thinking of an 
architecture for a location transparency mechanism that 
will be using the X.500 to retrieve presentationAddresses 
of management agents in a management platform.

Roughly speaking I want to be able to identify the 
presentationAddress of a management agent (e.g. SNMP 
agent) that contains management information for some 
network element. (By the term "network element" I mean 
equipment attached to the network (e.g. a router, 
a gateway, a workstation) ).  
Clearly, what I need first is a global, unique and stable 
naming schema for registering network elements in the DIT. 

Bearing in mind the "Charting Networks in the Directory" 
(OSI-DS-37) draft document I thought that maybe a refinement 
of the mechenism described there could be used for globally 
naming network elements within the Directory.
Of course the above naming schema results in machine names 
that are far away from user-friendly.  

As Steve writes:
>The decisions on naming machines should be primarily dictated by
> 1) Reasonable names for the machines
> 2) A naming structure which permits effective allocation.
>I would expect that some key services would be named at the ARC level,
>and the majority of machines at the departmental or project level
>(i.e., org units within ARC).  

I agree with these two points.
But for 2) I was expecting something related to the OSI-DS-37 idea.
What about the "nodes" mentioned there? Are we going to have finally
two places for registering machines (one under the OU level 
and the other under the network level)? Don't you think that all kinds 
of information about a machine must be gathered in one place? 

Although the more "natural" name for a machine under my organizational 
unit is e.g. host, csi, forth, gr  I would expect a seperate subtree for 
the machines within an organization. Although I could have 1,000 people 
entries mixed with 1,000 machine entries under the same organizational 
unit subtree a more efficient design is needed (that is going to be
transparent to the user, of course).

I think that a "network" subtree under the organizationalUnit level 
could be used for registering network elements. This provides also
a uniform view for a machine (= a part of the local network).  
With a search operation under the "network" subtree a DUA could find
information about a machine under the OU=ICS,O=FORTH,C=GR when the
given friendly name is host, ics, forth, gr.

I am looking forward to seeing opinions on the above.

Regards,
Costas.