Re: [midcom] RE: Last Call: 'Definitions of Managed Objects forMiddlebox Communication' to Proposed Standard (draft-ietf-midcom-mib)

Juergen Quittek <> Tue, 26 September 2006 09:05 UTC

Received: from [] ( by with esmtp (Exim 4.43) id 1GS8si-0004kP-DY; Tue, 26 Sep 2006 05:05:36 -0400
Received: from [] ( by with esmtp (Exim 4.43) id 1GS8sh-0004kJ-Cj for; Tue, 26 Sep 2006 05:05:35 -0400
Received: from ([]) by with esmtp (Exim 4.43) id 1GS8sd-0006TW-LN for; Tue, 26 Sep 2006 05:05:35 -0400
Received: from [] ( []) by (Postfix) with ESMTP id 533101BACA1; Tue, 26 Sep 2006 11:05:30 +0200 (CEST)
Date: Tue, 26 Sep 2006 11:05:26 +0200
From: Juergen Quittek <>
To: "Romascanu, Dan (Dan)" <>
Subject: Re: [midcom] RE: Last Call: 'Definitions of Managed Objects forMiddlebox Communication' to Proposed Standard (draft-ietf-midcom-mib)
Message-ID: <0ADCFB1A06A28CFC6ECA00AE@[]>
In-Reply-To: <>
References: <>
X-Mailer: Mulberry/3.1.6 (Mac OS X)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
X-Spam-Score: 0.0 (/)
X-Scan-Signature: 6a45e05c1e4343200aa6b327df2c43fc
Cc:, Martin Stiemerling <>, Pyda Srisuresh <>
X-Mailman-Version: 2.1.5
Precedence: list
List-Unsubscribe: <>, <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>

Dear Dan,

We are very sorry for the late reply on your comments and on Suresh's
replies.  The reason is that we (Martin and Juergen) wanted to discuss
them and did not find time to do so until today, because of our vacation
and business trip schedules.

Please find our replies inline.

If you are fine with them, we will post a new version of the I-D.


   Martin and Juergen

--On 28.08.2006 8:00 Uhr +0200 Pyda Srisuresh wrote:

> Dan,
> Thank you for your detailed comments. Sorry for the delayed response. I was
> away on vacation and just got back. Please see my responses below inline.
> regards,
> suresh
> --- "Romascanu, Dan (Dan)" <> wrote:
>> 1. Is the 'strict' SNMP terminology intentionally avoided in Section 4.2
>> and associated diagrams, and why? Meaning why do we say 'SNMP get
>> message' instead of 'SNMP GetRequest PDU', etc. ?
> [suresh] The objective of section 4.2 was essentially to indicate how the
> MIDCOM transactions can be mapped to SNMP. I.e., make the description of
> MIDCOM transactions to SNMP mapping easy to underdtand. Terms like "SNMP get
> message" and "SNMP put message" are simply easier for the reader to relate
> while talking about transactions.
>> 2. Section 5.3.1
>> > The MIDCOM MIB module does not require a middlebox to implement
>>    further specific MIB modules for supported middlebox functions, such
>>    as, for example, the NAT MIB module [RFC4008].
>> This should probably be 'further specific middlebox (NAT devices,
>> firewall) MIB modules'
>>    as, for example, the NAT MIB module [RFC4008].
> [suresh] OK; with a minor tweak to your suggested text as follows.
> s/"(NAT devices, firewall)"/"(NAT, Firewll)/
>> 3. Object midcomRuleAdminStatus
>> >     midcomRuleAdminStatus OBJECT-TYPE
>>        SYNTAX      INTEGER {
>>                        reserve(1),
>>                        enable(2),
>>                        notSet(3)
>>                    }
>> ...
>>  When retrieved, the object returns the last set value. If
>>             no value has been set, it returns one of the two possible
>>             values."
>>        DEFVAL { notSet }
>> I do not understand what are the 'two possible values'. What happens of
>> a retrieval happens before the object is set for the first time?
> [suresh] Oops... Will change the text to read as follows.
> When retrieved, the object returns the last set value. If
> no value has been set, it returns the default value of notset(3).
>> 4. Several DESCRIPTION clauses (e.g. midcomRuleAdminStatus,
>> midcomRuleStorageType) include SNMP-specific error messages when
>> describing the behavior of the object. This is OK, as the MIDCOM-MIB is
>> designed to be used with SNMP as MIDCOM protocol, yet I would include a
>> note on this subject because this is not customary within other MIB
>> documents which are written with a protocol-independent orientation.
> [suresh] I will leave to Juergen or Martin to comment on this.

We assume that you refer to error code 'inconsistentValue' that we
mention in several DESCRIPTION clauses.

We think this is still customary in recent MIB modules from where we
got the idea to use this error code.  See, for example, RFCs 4001,
4087, 4131, 4149, 4268, 4368, 4444, and 4546.

>> 5. What happens with the object midcomRuleObjectTime when
>> midcomRuleObjectType is permanent(4)? The DESCRIPTION clause of the type
>> object suggests that time is read-only. With DEFVAL 0 this means
>> automatic destruction of the row at the end of the transaction. Is this
>> the desired behavior, or did I mis-understand something?
> [suresh] Yes, I believe, that is the desired behaviour.

This is a good catch.  We did not consider storage type 'permanent'
well when writing the DESCRIPTION clause.

If the storage time is permanent, the value should never become 0.
In this case, we suggest this object to have a permanent value of 4294967295.

We propose appending the following paragraph to the DESCRIPTION clause:

        "If object midcomRuleStorageType indicates that the policy
         rule has storage type permanent(4), then this object has
         a constant value of 4294967295."

> Note, however, the DESCRIPTION for midcomRuleStorageType says that attempts to
> set this object to permanent will always fail with an inconsistentValue error.
> And, the default value for this object is volatile(2).
>> 6. I do not feel comfortable with the DESCRIPTION clause of
>> midcomRuleError RECOMMENDing values for this object without defining
>> what behavior each message is supposed to cover. This type of object is
>> not interoperable, and this would be made clear if it was said that
>> these are examples rather than RECOMMENDations.
> [suresh] I am OK with listing the error strings as examples, rather than as
> recommendations. I will leave to Juergen or Martin to further comment on this.

The errors a re well defined in the MIDCOM protocol semantics (RFC 3989,
sections 2.3.9 and 2.3.10).
We suggest appending to the DESCRIPTION clause:

        "The semantics of these error messages and the corresponding
         behavior of the MIDCOM MIB implementation are specified
         in sections 2.3.9 and 2.3.10 of RFC 3989."

and adding a reference clause

        "RFC 3989, sections 2.3.9 and 2.3.10"

>> 7. Another side-effect of the midcomRuleObjectType being permanent(4) is
>> that the permanent rules cannot be applied to interfaces, so they can be
>> only global. Same about transport protocol and other read-write objects.
> [suresh] As I said earlier, attempts to set midcomRuleStorageType to permanent
> will always fail with an inconsistentValue error.

As stated in the DESCRIPTION clause of midcomRuleStorageType, a permanent(4)
row has all write access to other objects in the row disabled.

>> 8 . There is no DEFVAL for midcomRuleFlowDirection
> [suresh] Right, that was intentional.

We added

     DEFVAL { outbound }

as Suresh had suggested in another email.

>> 9.
>> >          Valid values for midcomRuleTransportProtocol
>>             other than zero are defined at:
>> Does this need a new type of registry from IANA? There is nothing in the
>> IANA considerations about this.
> [suresh] Well, nothing specific to MIDCOM MIB per se. IANA assigns IP protocol
> numbers independently, right.
>> 10. What notification needs to be sent when midcomConfigIfEnabled is set
>> to false? Neither the DESCRIPTION of the object nor the one of the
>> notifications do provide any clue.
> [suresh] The DESCRIPTION for midcomConfigIfEnabled does say the following.
>    Setting
>    this object to false(2) immediately stops middlebox
>    support at the indexed IP interface.  This implies that
>    all policy rules that use NAT or firewall resources at
>    the indexed IP interface are terminated immediately.
>    In this case, The MIDCOM agent MUST send notifications
>    to all MIDCOM clients that have access to one of the
>    terminated rules.
> As for the rule termination event, please refer section 7.9.

The notifications to be sent are of type midcomUnsolicitedRuleEvent.
Suggestion: s/notifications/midcomUnsolicitedRuleEvent


   Martin and Juergen

> regards,
> suresh
>> > -----Original Message-----
>> > From: The IESG []
>> > Sent: Thursday, August 03, 2006 8:09 PM
>> > To: IETF-Announce
>> > Cc:
>> > Subject: Last Call: 'Definitions of Managed Objects for
>> > Middlebox Communication' to Proposed Standard (draft-ietf-midcom-mib)
>> >
>> > The IESG has received a request from the Middlebox
>> > Communication WG to consider  the following document:
>> >
>> > - 'Definitions of Managed Objects for Middlebox Communication '
>> >    <draft-ietf-midcom-mib-08.txt> as a Proposed Standard
>> >
>> > The IESG plans to make a decision in the next few weeks, and
>> > solicits final comments on this action.  Please send any
>> > comments to the or mailing lists
>> > by 2006-08-17.
>> >
>> > The file can be obtained via
>> >
>> >
>> >
>> > _______________________________________________
>> > IETF-Announce mailing list
>> >
>> >
>> >
>> _______________________________________________
>> midcom mailing list
> _______________________________________________
> midcom mailing list

midcom mailing list