Re: [Lsr] Moving Forward [Re: Flooding Reduction Draft Redux]

[Peter Psenak <[e-mail]>]

Hi Huaimo,

On 18/02/2019 16:28 , Huaimo Chen wrote:
> Hi Peter,
>
>> -----Original Message-----
>> From: Peter Psenak [mailto:[e-mail]]
>> Sent: Thursday, February 14, 2019 2:30 AM
>> To: Huaimo Chen <[e-mail]>; Acee Lindem (acee) <[e-mail]>; Christian Hopps <[e-mail]>; [e-mail]
>> Subject: Re: [Lsr] Moving Forward [Re: Flooding Reduction Draft Redux]
>>
>> Hi Huaimo,
>>
>> On 13/02/2019 22:50 , Huaimo Chen wrote:
>>> Hi Peter,
>>>
>>>     My explanations/answers are in line below with prefix [HC].
>>>
>>> -----Original Message-----
>>> From: Peter Psenak [mailto:[e-mail]]
>>> Sent: Wednesday, February 6, 2019 4:58 AM
>>> To: Huaimo Chen <[e-mail]>; Acee Lindem (acee)
>>> <[e-mail]>; Christian Hopps <[e-mail]>; [e-mail]
>>> Subject: Re: [Lsr] Moving Forward [Re: Flooding Reduction Draft Redux]
>>>
>>> Hi Huaimo,
>>>
>>> On 03/02/2019 17:58 , Huaimo Chen wrote:
>>>> Hi Acee,
>>>>
>>>>
>>>>
>>>>     I agree with you on keeping the signaling for two modes. The
>>>> other parts for the distributed solution need to be removed.
>>
>> optimized flooding is not only about algorithm to calculate the flooding topology and the way
>> it is distributed/computed. It is also about local rules to make sure the flooding remains consistent.
>> These are _independent_ of centralized/distributed modes. And it make no sense to specify
>> these rules in two drafts.
>
> [HC] The rules/procedures/behaviors that are common to both the centralized and distributed
> solution/mode should be in one document. These common parts will be used by both solutions/modes.
> They are described in the two drafts and should be merged based on technical merits.
>
> In addition to the common parts, there are still some rules/procedures/behaviors that are
> different from one mode to another mode. For example, the rule/procedure for fault tolerance
> to failures used in the centralized solution/mode will be different from that used in the distributed
> solution/mode.

I don't think there is any significant difference needed. And I believe 
these local rules for both modes should reside in a single document as 
most of them are applicable to both modes.

> In the distributed solution/mode, there will be a set of rules/procedures/behaviors
> that are common to the algorithms for computing flooding topology and specific to the
> distributed solution/mode. For example, a specific procedure for scheduling an algorithm
> to compute a flooding topology will be used on every node for the distributed solution/mode.

scheduling a distributed versus centralized computation is not something 
that makes the behavior different. Sure, you can put a scheduling 
details for a particular algorithm in the document that describes the 
distributed algorithm itself.

thanks,
Peter



>
>>>
>>> There are no "other" parts specific for the distributed solution.
>>>
>>> [HC] Some behaviors for the distributed solution/mode are described in draft-li-dynamic-flooding.
>>> For example, there are a few of places from page 27 to 30, which define the behaviors specific
>>> for the distributed solution/mode.
>>
>> I strongly disagree. The fact that we say in centralized mode area leader recomputes and in distributed
>> mode all nodes recompute make no difference in behavior.
>
> [HC] It seems better for some of them to be more general.
>
> Best Regards,
> Huaimo
>
>> thanks,
>> Peter
> .
>