Re: [CCAMP] Vendor-Specific Application Code in draft-ietf-ccamp-rwa-wson-encode

[Leeyoung <[e-mail]>]

Hi Gert,

Yes, you hit the nail on the head. Thanks.

It looks like we can move on this issue with the latest development in this thread.

Best regards,
Young

-----Original Message-----
From: Gert Grammel [mailto:[e-mail]] 
Sent: Thursday, January 29, 2015 10:00 AM
To: [e-mail]; Leeyoung; 'Varma, Eve L (Eve)'; [e-mail]; 'Lam, Hing-Kam (Kam)'; [e-mail]
Cc: [e-mail]; [e-mail]; [e-mail]; [e-mail]
Subject: RE: [CCAMP] Vendor-Specific Application Code in draft-ietf-ccamp-rwa-wson-encode


Hi Leeyoung, 

Keep in mind that G.698.2 allows to use transceivers from different vendors, based on standard AIs (=Application Codes). Now in case of a multi-vendor network, *additional* proprietary application codes can be used if there is a pre-knowledge that the proprietary AIs match. I believe this was one case you were referring to.
So a single transceiver HW implementation supporting the standard-AI will almost certainly support also a vendor-AI e.g.,  if a higher level of OSNR can be tolerated than the standard-AI has encoded. Hence we can conclude that a single (vendor) transceiver potentially supports a list of AIs. So even by picking Option 2 your case can be addressed.

How to encode those AIs is another matter as printable strings are not always the best choice to use. https://tools.ietf.org/html/draft-martinelli-wson-interface-class-03 and https://tools.ietf.org/html/draft-dharini-netmod-g-698-2-yang-02 already provide some encoding, whereby OUI would need to be worked in.

Gert


-----Original Message-----
From: Adrian Farrel [mailto:[e-mail]] 
Sent: 29 January 2015 14:24
To: 'Leeyoung'; 'Varma, Eve L (Eve)'; [e-mail]; 'Lam, Hing-Kam (Kam)'; Gert Grammel; [e-mail]
Cc: [e-mail]; [e-mail]; [e-mail]; [e-mail]
Subject: RE: [CCAMP] Vendor-Specific Application Code in draft-ietf-ccamp-rwa-wson-encode

Hi again,

> There is always a priori knowledge in optical network domain as to who 
> are you interfacing with. So you know which vendor you are interfacing.
> If you do not know, then you are in trouble.

Hmmm. It is exactly type of trouble we are trying to detect and protect against.

I refute your statement of a priori knowledge. I think there is a priori intention, but not knowledge. Unless you have very good eyesight or someone at the other end of the fiber when you give it a tug, you don't know. And even then. Fibering errors happen from time to time. Consider, in particular a patch panel.

> Now, what is the purpose of standard FECs and modulations in the AI? 
> Given several choices each vendor may support in its device, the path 
> computation would find a matched types for FEC and modulation for a given optical path.
> This is what is intended when optical signal processing constraints 
> were proposed as part of path computation constraints in optical networks.


The case you are making here is for no standard control plane!
What is the point of standardising if there is never any interworking?
But actually, we know about interworking at the physical layer, and (more important) we know about a single, end-to-end control plane that spans multiple vendor devices. It all exists.

Of course, we can fall back into the old-style vendor islands, and many like to do so. But it is not a compulsory deployment model.

> There is very little chance for vendor specific FECs and Modulations 
> will match even if they are identified with the OUI code.

You have it the wrong way round!
The OUI is largely to protect against expectations of interworking when none can exist.
It might (much less frequently) be used to describe the way that vendorA and vendorB pick FECs and modulations in order to achieve interworking.

Adrian