Re: [IPFIX] Questions about compliance statement in IPFIX-selection-techniqes draft

[Paul Aitken <[e-mail]>]

Benoit, Nevil, All,

Option 3, "yes and no".

A clear specification of the compliance criteria is a must.

However, with the current compliance statement an implementation of all 
the specified Flow Sampling techniques without Property Match Filtering, 
is not compliant. Similarly, a device which implements Hash-based Flow 
Filtering rather than Property Match Filtering is not compliant.

We can't predict what future device requirements will be or how the 
draft will be used. Therefore the choice of Property Match Filtering as 
the compliance criteria seems somewhat arbitrary. It may seem like an 
obvious choice today, but in future might turn out like the 640KB limit. 
We don't want this to be a decision we regret with hindsight as we 
struggle with an awkward implementation of Property Match 
Filteringsimply to claim compliance.

So I'd prefer the compliance statement to say, "In order to be compliant 
with this document, at least one of the Filtering or Sampling techniques 
MUST be implemented."
Or, "... at least one of the Intermediate Flow Selection Process 
Techniques specified in section 6 MUST be implemented".

P.


On 22/05/13 07:35, Benoit Claise wrote:
> Hi Nevil,
>
> "yes, keep the compliance statement in section 6.1".
> I want to keep the consistency to makes it clear to implementors 
> what's needed for a minimal conforming
> implementation, and to be similar to RFC 5475
>
> Regards, Benoit (as a contributor)
>> Hi all:
>>
>> draft-ietf-ipfix-flow-selection-tech-16 is now in IESG Evaluation;
>> it currently has a blocking DISCUSS
>>
>> Section 6.1 contains a compliance statement, which says
>>   "In order to be compliant with this document, at least the
>>    Property Match Filtering MUST be implemented."
>> The AD concerned asks whether we could remove this requirement.
>>
>> This was triggered by a remark in my Shepherd statement for this draft,
>> which said
>>   "This draft raised IPR concerns, in the same manner as the PSAMP
>>   selection draft had done.  Nick Duffield (AT&T) commented that
>>   the AT&T IPR claim relates only to statistical sampling, and PSAMP
>>   handled this by saying "at least on of the sampling techniques
>>   must be implemented."
>>   In this draft, we have tightened that up a little by saying
>>   "a conforming implementation MUST implement at least the
>>   Property Match Filtering."
>> I realise now that the IPR in question was that relating to
>> draft-krishnan-opsawg-large-flow-load-balancing, which was discussed
>> on the IPFIX list back in Feb/Mar 2013; the issue that raised was
>> cleared by the company concerned making an IPR declaration.
>>
>> That leaves us with the question about the compliance requirement.
>> The motivation for this remains the same as it was for RFC 5475,
>> "Sampling and Filtering Techniques for IP Packet Selection,"
>> the statement in section 6.1 simply refines it a little, and makes it
>> clear to implementors what's needed for a minimal conforming
>> implementation.
>>
>> So now, I need some feedback from the WG.  Please send a note to
>> the list saying
>> "yes, keep the compliance statement in section 6.1" or
>> "no, delete the compliance statement in section 6.1.
>>
>> Real soon now, please!
>>
>> Cheers, Nevil
>>
>>
>>
>
> _______________________________________________
> IPFIX mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/ipfix