Re: [pkt-prov] MTA MIB Error OID table - a number of questions

Hi -

> From: "Eugene Nechamkin" <[email protected]>
> To: "Randy Presuhn" <[email protected]>
> Cc: <[email protected]>
> Sent: Wednesday, January 12, 2005 8:44 PM
> Subject: RE: [ipcdn] [pkt-prov] MTA MIB Error OID table - a number of questions
>
> If your concern is the usage of the object in the different lingual
> environments with presumably normative language in the object's
> description, then the proposed, I believe, adequately addresses your
> concern. If your concern is the interoperability, then the object
> itself, without the normative requirement, carries the informative
> meaning for the NMS personal, and is becoming even wider interoperable.
>
> If this is still not satisafctory, please feel free to make a specific
> editorial suggestion.
...

A managed device has no way of knowing what locale(s) may be in effect
on the network management systems (both human-controlled and purely
automated) that are communicating with it.  The locale in effect may even
change, depending on which operator is currently using a management system.
Consequently, localization of error messages is something to be done at the
management systems, not the managed devices.

NMS operators are not all human.  More and more work is done by scripts
and programs working on behalf of humans.  In general software isn't all
that good at figuring out "semantic equivalents".  The problem is even
worse if the localized text doesn't carry some kind of language tag.

Error messages are quite different from names and labels.  The latter can
be treated as opaque values by software, and in general there's no a priori
value that can be associated with them.  Error messages, on the other hand,
do have semantics, and management software will be expected to alter its
behaviour based on the messages it encounters.

My preference is for a machine-readable enumeration of error codes.
If that is not practical, then, at least for the cases that a script or automated
management system would need to recognize, the EXACT strings
that appear in protocol should be spelled out, with recognition that
a user interface might present some localized equivalent.

Randy