Re: Re: MIBDoctorreview:publishdraft-ietf-disman-remops-mib-v2-06.txt

Juergen Schoenwaelder <[email protected]> Mon, 4 Jul 2005 22:01:26 +0200
Newsgroups gmane.ietf.disman
Message-ID <[email protected]>
On Mon, Jul 04, 2005 at 12:10:08PM -0700, Randy Presuhn wrote:
> ...
> > Basically, it is showing the actual name/address resolution at
> > the remote device.  But should it be just what DNS delivers to the
> > device or should the local configuration of name resolution be included?
> 
> As a technical contributor...
> 
> The operation is described in terms of the gethostbyname()
> and gethostbyaddr() APIs, rather than protocol interactions.
> Consequently, my interpretation is to use whatever those APIs
> (or their equivalents) return on that system.  Whether the APIs
> invoke DNS, NIS, a local database, something else, or some
> combination isn't a consideration.
> 
> If this MIB is being used to diagnose application outages, this
> seems a reasonable interpretation.  I believe our thinking was
> that if the objective was to diagnose problems with the DNS
> infrastructure itself, one would employ RFC 1611 / 1612.
> 
> Those have been declared Historic, so perhaps that rationale
> might no longer be as sensible.  However, given where we are
> in the process with this document, I think we should leave things
> as they are: the table is populated with results from
> gethostbyname() and gethostbyaddr().  If this isn't sufficiently
> clear in the i-d, then I wouldn't object to adding a clarification.

The results table uses the InetAddress convention to return the 
results. For DNS names, the InetAddressType would be dns(16). For 
other names, unknown(0) seems applicable.  The unfortunate news 
is that without some effort, there is no way to decide whether 
the name returned by gethostbyaddr() is a DNS name or not.

Another question concerns dual hosts with multiple names for different
addresses which may actually life in different address domains. Can an
application just pick randomly one address, call gethostbyaddr() or 
getnameinfo() and report the result? Or is it required to collect
and consolidate the information for all IP addresses? The current ID
uses language such as "official name" or "primary address" which are
not well defined.

/js

-- 
Juergen Schoenwaelder		    International University Bremen
<http://www.eecs.iu-bremen.de/>	    P.O. Box 750 561, 28725 Bremen, Germany