RE: partial comments on section 5 of ibif

Ok,

Now I REALLY understand what Randy is saying...

I think the answer is that we will specify that we will report 64 bit
octet/packet counters as is required.  We will also specify the
multicast counters.

If an implementation is not capable of reporting these values out - the
implementation can return NOSUCHINSTANCE from snmpv2 and beyond or
NOSUCHOBJECT for v1.  

What are the objections to this - can you please get your objections to
this by the end of Last Call, or this will be concensus.

Bill

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf
Of Randy Presuhn
Sent: Thursday, October 09, 2003 11:10 PM
To: [email protected]
Subject: Re: [Ipoverib] partial comments on section 5 of ibif


Hi -

> From: "bill" <[email protected]>
> To: "'Randy Presuhn'" <[email protected]>
> Cc: <[email protected]>
> Sent: Thursday, October 09, 2003 10:31 PM
> Subject: RE: [Ipoverib] partial comments on section 5 of ibif
>

> Randy, (let me take my WG Chair hat off)
>
> I don't want to disagree with your opinion of many of the interesting 
> objects that are giving us fits (multicast packet counting)
>
> Let me try to change specification and implementation issues - we 
> should specify multicast packets in/out whatever.  Then it is up to 
> the implementation to provide a value for the counter - not implement 
> the object (NO SUCH INSTANCE, NO SUCH OBJECT for V1)

so far, I am in total agreement with you.

>                                                  or return 0 - on an

No, returning a constant zero is *NOT* an option.  If multicast can
happen, then the choice is between implementing a counter that actually
counts, or not implementing it at all.  A counter that doesn't count the
events it is supposed to count is a bug, with the unfortunate potential
of misleading management systems to draw false conclusions about the
network's operation or performance.

> implementation decision.  I realize the IETF does not specify 
> implementations, or provide conformance testing (This implementations 
> fully supports RFC X)

You're right that we don't do the actual conformance tests. However, we
*do* define what it means to conform to a MIB module. That's what RFC
2580  is all about.

> Your objections are that we are specifying implementation behavior, 
> rather than what an ideal specification.  It is then up to the 
> implementations to implement these counters in some way
...

Agreed.  Thanks for restating my point so clearly.

Randy



_______________________________________________
IPoverIB mailing list
[email protected] https://www1.ietf.org/mailman/listinfo/ipoverib