Re: [CCAMP] [RTG-DIR] RtgDir review: draft-ietf-ccamp-general-constraint-encode-16.txt

[Lou Berger <[e-mail]>]

Young,
    I think we're close, but lets tighten up the language a bit.


On 2/3/2015 12:08 PM, Leeyoung wrote:
> Lou,
>
> OK, How about:
>
> OLD 
>
> ... There may be more than one matrix associated with
>    a node as the node can partition the switch matrix into several sub-
>    matrices for various reasons such as incremental updates, etc. When
>    the matrix is partitioned into sub-matrices, it is envisioned that
>    they are mutually exclusive to one another in representing which
>    ports/labels are associated with each sub-matrix. This implies that
>    two matrices will not have the same {src port, src label, dst port,
>    dst label}.
>
>    Each sub-matrix is assigned a unique Matrix ID to represent a 
>    mutually exclusive set of {src port, src label, dst port, dst label} 
>    from other sub-matrices.
>
> NEW: 
>
> ... There may be more than one matrix associated with
>    a node as the node can partition the switch matrix into several sub-
>    matrices for various reasons primarily to limit the size of any 
>    individual information element used to represent the matrix or allow 
>    incremental updates, etc. When the matrix is partitioned into 
>    sub-matrices, it is envisioned that they are mutually exclusive to 
>    one another in representing which ports/labels are associated with 
>    each sub-matrix. This implies that two matrices will not have the 
>    same {src port, src label, dst port, dst label}.
>
>    Each sub-matrix is assigned a unique Matrix ID to represent a 
>    mutually exclusive set of {src port, src label, dst port, dst label} 
>    from other sub-matrices.

How about:

   There may be more than one Field associated with
   a node as a node can partition the switch matrix into several sub-
   matrices. This partitioning is primarily to limit the size of any 
   individual information element used to represent the matrix and to 
   enable incremental updates. When the matrix is partitioned into 
   sub-matrices, each sub-matrix will be mutually exclusive to 
   one another in representing which ports/labels are associated with 
   each sub-matrix. This implies that two matrices will not have the 
   same {src port, src label, dst port, dst label}.

   Each sub-matrix is identified via a different Matrix ID which MUST 
   represent a unique combination of {src port, src label, dst port, 
   dst label}.


Lou



> Would this be Ok with you? Then I will publish a revision upon your approval.
Also we need to hear from Tomonori on if his issues is addressed.

Thanks,
Lou
> Thanks,
> Young
>
> -----Original Message-----
> From: Lou Berger [mailto:[e-mail]] 
> Sent: Tuesday, February 03, 2015 7:59 AM
> To: Leeyoung
> Cc: '[e-mail]'; '[e-mail]'; '[e-mail]'; Tomonori Takeda; [e-mail]
> Subject: Re: [RTG-DIR] RtgDir review: draft-ietf-ccamp-general-constraint-encode-16.txt
>
> Young,
>
> I guess I didn't see the new text as substantively different or more
> informative than preceding text:
>
>    ... There may be more than one matrix associated with
>    a node as the node can partition the switch matrix into several sub-
>    matrices for various reasons such as incremental updates, etc. When
>    the matrix is partitioned into sub-matrices, it is envisioned that
>    they are mutually exclusive to one another in representing which
>    ports/labels are associated with each sub-matrix. This implies that
>    two matrices will not have the same {src port, src label, dst port,
>    dst label}.
>
> I'd suggest directly addressing the ambiguity by revising the first
> sentence and "for various reasons such as incremental updates, etc. " in
> particular.  Perhaps "primarily to limit the size of any individual
> information element used to represent the matrix."  or however else
> you'd like to capture the intent.
>
> Lou
>
> On 2/2/2015 5:15 PM, Leeyoung wrote:
>> Hi Lou,
>>
>> The original question Tomonori raised was if sub-matrices are like virtual nodes attached to connectivity matrix. The answer was no, and sub-matrix is simply a partition of connectivity matrix into smaller pieces. To that end, I added the following text in Section 2.1
>>
>>    Each sub-matrix is assigned a unique Matrix ID to represent a 
>>    mutually exclusive set of {src port, src label, dst port, dst label} 
>>    from other sub-matrices. 
>>
>> Let me know what you think.
>>
>> Thanks,
>> Young
>>
>> -----Original Message-----
>> From: Lou Berger [mailto:[e-mail]] 
>> Sent: Monday, February 02, 2015 3:29 PM
>> To: Leeyoung
>> Cc: Tomonori Takeda; [e-mail]; '[e-mail]'; '[e-mail]'; '[e-mail]'
>> Subject: Re: [RTG-DIR] RtgDir review: draft-ietf-ccamp-general-constraint-encode-16.txt
>>
>> Hi Young,
>>
>> I see you just published an update.  I have one question below.
>>
>> On 1/23/2015 3:01 AM, Tomonori Takeda wrote:
>>> Hi Young,
>>>
>>> OK, thanks,
>>>
>>> Tomonori
>>>
>>> -----Original Message-----
>>> From: Leeyoung [mailto:[e-mail]] 
>>> Sent: Thursday, January 22, 2015 2:44 AM
>>> To: Tomonori Takeda（武田知典）; Leeyoung; [e-mail]
>>> Cc: '[e-mail]'; '[e-mail]'; '[e-mail]'
>>> Subject: RE: [RTG-DIR] RtgDir review: draft-ietf-ccamp-general-constraint-encode-16.txt
>>>
>>> Hi Tomonori,
>>>
>>> Thanks for your comment. Please see in-line for my response. Please let me know if the response would satisfy you. 
>>>
>>> Best regards,
>>> Young
>>>
>>> -----Original Message-----
>>> From: Tomonori Takeda [mailto:[e-mail]] 
>>> Sent: Wednesday, January 21, 2015 12:48 AM
>>> To: Leeyoung; [e-mail]
>>> Cc: '[e-mail]'; '[e-mail]'; '[e-mail]'; Tomonori Takeda
>>> Subject: RE: [RTG-DIR] RtgDir review: draft-ietf-ccamp-general-constraint-encode-16.txt
>>>
>>> Hi Young,
>>>
>>> Thanks.
>>>
>>> Two follow-up questions/comments.
>>> (I am fine with other points, which you already addressed in the updated draft.)
>>>
>>>> 2) In section 2.1, it says "two matrices will not have the same {src port, src label, dst port, dst label}". To be precise, I guess this should be > "two matrices will not have the same {src port, src label}, and two matrices will not have the same {dst port, dst label}"?
>>>>
>>>> YOUNG>> I think your suggestion may be too restrictive. For instance, if we have one source (port 1) and one destination (port 2) with two labels > each. Then we would have: {(1,1,2,1), (1,1,2,2), (1,2,2,1), (1,2,2,2)} I think with the current statement, we can send this info in any combination > of multiple matrices, which I think perfectly fine. With your suggestion, I would not be able send (1,1,2,1) and (1,1,2,2) together. Why would this > not be made possible? My take is as long as each submatrix represents a set of disjoint quadruples, that should be allowed.
>>> My reading of "two matrices will not have the same {src port, src label, dst port, dst label}" is as follows.
>>>
>>> <Example A>
>>>
>>>   input port=1  --> Submatrix#1 --> output port=2
>>>   input label=1                     output label=1
>>>
>>>   input port=1  --> Submatrix#2 --> output port=2
>>>   input label=1                     output label=2
>>>
>>>   This is allowed.
>>>
>>> <Example B>
>>>
>>>   input port=1  --> Submatrix#1 --> output port=2
>>>   input label=1                     output label=1
>>>
>>>   input port=1  --> Submatrix#2 --> output port=2
>>>   input label=1                     output label=1
>>>
>>>   This is not allowed.
>>>
>>> <Example C>
>>>
>>>   input port=1  --> Submatrix#1 --> output port=2
>>>   input label=1                     output label=1
>>>
>>>   input port=1  --> Submatrix#2 --> output port=2
>>>   input label=2                     output label=2
>>>
>>>   This is allowed.
>>>
>>> Is above understanding correct?
>>> If so, I am not sure how example A works, since I am not sure what is the indentifier to direct from input to each submatrix.
>>>
>>> Maybe I am mis-understanding what sub-matrix is. I thought sub-matrix is a sort of virtual node, splitting the single matrix (or switch) into smaller pieces.
>> Can you explain how the revised text clarifies this point?
>>
>> Thanks,
>> Lou
>>
>>> ....
>