It's hard to imagine any good way to implement the Config Changes object,
since the usual SNMP agent implementation techniques don't provide any good
way to know whether any of a bunch of other objects have changed, short of
the agent's polling all of them whenever Config Changes is queried. Is
this what everybody is doing to implement this?
Actually some informal testing here suggests that what many of us printer
vendors are really doing with this is nothing, i.e. not implementing the
Config Changes object, which wouldn't be very surprising given the
difficulty mentioned above.
This object would appear to be considerably worse than useless, i.e.
potentially very dangerously misleading, if it is not implemented (since it
purports to be mandatory), and implemented uniformly. Per the point above,
it seems often (and understandably) to be unimplemented. And this current
discussion, proving the current uncertainty about which objects are
considered part of the configuration, establishes that any current
implementations which do exist are certainly non-uniform in that respect.
To summarize, Config Changes seems to be intolerably non-uniform at two
levels, namely whether it's implemented at all, and, if so, what's included.
It seems that before any client/manager could ever dare use the Config
Changes object, it must first have some way to be confident whether the
agent implements the object, and what objects are considered to be included
in the configuration.
Shouldn't we either provide these things - or else just admit that Config
Changes is a lost cause?
Date: Tue, 4 Aug 1998 20:14:11 PDT
From: Gail Songer <email@example.com>
Subject: Re: PMP> Re: COnfig Changes
As keeper of the FAQ I am more than happy to add these variables
and have a new version ready to go as soon I hear the word.
The only comment that I have is that this was not intended to be a
complete list but was to give a flavor of the things that could cause
a config change. Should this intent change? Do we want to state ALL
those things that an ideal implementation would count as config
changes? Should the list be an "only" list or an "at least" list?
Ira Mcdonald x10962 wrote:
> I agree with Harry and Jay that all of these MIB variables should
> be added to what constitutes a config change
> - Ira McDonald (High North Inc)
> [Harry's note]
> From: Harry Lewis <firstname.lastname@example.org>
> To: <email@example.com>
> Subject: PMP> COnfig Changes
> Message-Id: <5030100024028591000002L012*@MHS>
> Date: Mon, 3 Aug 1998 09:58:39 PDT
> Mime-Version: 1.0
> Content-Type: text/plain; charset=iso-8859-1
> Content-Transfer-Encoding: quoted-printable
> Sender: firstname.lastname@example.org
> Status: R
> There is a FAQ (#7) - that explains what constitutes a config change.
> We have some questions...
> 1) The following variables weren't included:
> This seems inconsistent with other listed 'default' settings --
> prtChannelDefaultPageDescLangIndex, prtInterpreterDefaultOrientation,
> prtInterpreterDefaultCharSetIn, prtInterpreterDefaultCharSetOut.
> 2) These variables were also not included but could be considered chang=
> to the configuration although they do not effect operations.
> Harry Lewis - IBM Printing Systems