RE: Optical Link Interface

[Jonathan Lang <[e-mail]>]

Osama,
  Please see inline comments.
 
Thanks,
Jonathan

-----Original Message-----
From: Osama Aboul-Magd [mailto:[e-mail]]
Sent: Wednesday, July 25, 2001 10:21 AM
To: Jonathan Lang; Andre Fredette; [e-mail]
Subject: RE: Optical Link Interface



Jonathan, 

Comments are inline. 

Regards; 

Osama Aboul-Magd 
Nortel Networks 
P.O. Box 3511, Station "C" 
Ottawa, ON, Canada 
K1Y - 4H7 
Tel: [phone] 
e.mail: [e-mail] 

 -----Original Message----- 
From:   Jonathan Lang [ mailto:[e-mail]
<mailto:[e-mail]> ] 
Sent:   Wednesday, July 25, 2001 11:53 AM 
To:     Aboul-Magd, Osama [CAR:1A00:EXCH]; Andre Fredette;
[e-mail] 
Subject:        RE: Optical Link Interface 

Osama, 
  Please see inline comments. 

Thanks, 
Jonathan 

>-----Original Message----- 
>From: Osama Aboul-Magd [ mailto:[e-mail]
<mailto:[e-mail]> ] 
>Sent: Wednesday, July 25, 2001 8:15 AM 
>To: Andre Fredette; [e-mail] 
>Subject: RE: Optical Link Interface 
> 
> 
>Andre, 
>Obviously I disagree with you in your conclusion. There are number of 
factors that should 
> be considered: 
>- While LMP has been in STANDARD process for some time, there hasn't been 
much deployment, 
> if any, of it. Furthermore the reality of optical networks today is that 
same vendor 
> equipment is deployed in a single span. Implementation of LMP in this case

is at best 
> optional. One may choose not to implement LMP at all. IMHO tying this 
important interface 
> to the fate of LMP is not the way to proceed. How long LMP been at the 
IETF is irrelevant. 
> - NTIP is a new TCP application in the same way as LMP-WDM is a NEW 
application for LMP. 
I'm glad you brought this up.  TCP greatly complicates the fault recovery 
procedure and fault recovery is an essential aspect for this interface. 

        [Osama] I am not sure I follow. 

> - The issue is not really LMP vs TCP. NTIP and LMP-WDM differ in the model

itself. Your 
> choice to use LMP to run LMP-WDM on top of it forced you to treat the PXC 
and LS as peers. 
> NTIP treats PXC-LS relationship as master-slave. Non peer relationship 
suits this 
> interface better. 
I understand this limitation works for your product, but other WDM vendors 
have requested a more flexible model. 

        [Osama] what is the limitation here? Are you saying having a simple
design is a limitation? Not everything has to be complex.  

[Jonathan]  Other DWDM vendors are not happy with the master-slave model.
Also, the claim that NTIP is simple is an explicit assertion and you seem to
be trying to make an implicit assertion that LMP is complex. 

> - To link with other standards bodies, as far as I know T1X1, where the 
optical knowledge 
> exists, is yet to arrive to a definition of this interface. More 
contributions are 
> solicited to better understand its need and applications. Given that, I'd 
like to 
> understand the rationale for the "rough consensus" reached at CCAMP 
mailing list. Only 4-5 
> messages were posted to the mailing list and mostly from authors of the 
OLI requirements 
> draft. 
> - I don't see harm in having two protocols. The market made its decisions 
on the examples > you mentioned, and I am sure it will happen again in this 
case. 
Given the CR-LDP fiasco in MPLS, it was clearly stated by the ADs and 
Working Group chairs in Minnesota that only one protocol will progress in 
the IETF. 
        
[Osama] I don't understand why CR-LDP and RSVP-TE have been brought to this
discussion. This is a completely different situation. LMP hasn't seen the
wide deployment that RSVP-TE has. The example is inadequate. 

[Jonathan] CR-LDP and RSVP-TE were both being developed prior to either one
of them being widely deployed.  The effort involved in developing 2
protocols concurrently to do the same thing is widely perceived as being
counter productive. 

> Regards; 
> Osama Aboul-Magd 
> Nortel Networks 
> P.O. Box 3511, Station "C" 
> Ottawa, ON, Canada 
> K1Y - 4H7 
> Tel: [phone] 
> e.mail: [e-mail] 
 -----Original Message----- 
From:   Andre Fredette [ mailto:[e-mail]
<mailto:[e-mail]> ] 
Sent:   Monday, July 23, 2001 10:01 PM 
To:     [e-mail] 
Subject:        Optical Link Interface 
Last month, the "Optical Link Interface (OLI) Requirements" document 
http://search.ietf.org/internet-drafts/draft-many-oli-reqts-00.txt
<http://search.ietf.org/internet-drafts/draft-many-oli-reqts-00.txt>  
was discussed on the ccamp mailing list and achieved "rough consensus" 
according to our working group co-chairs.  Given this, we'd like to make 
some progress on the protocol specification.  I'd like to generate some 
discussion on this mailing list before the IETF meeting in London because 
meaningful technical discussion cannot occur in the time allocated during 
the meeting. 
There have been two proposals in the IETF to satisfy the OLI requirements: 
1. [LMP-WDM]: 
"Link Management Protocol (LMP) for DWDM Optical Line Systems" 
http://www.photonex.com/other/draft-fredette-lmp-wdm-02.txt
<http://www.photonex.com/other/draft-fredette-lmp-wdm-02.txt>  
(note, this updated document was submitted Friday, so it should show up 
on the official website soon). 
and 
2. [NTIP]: 
"Network Transport Interface Protocol (NTIP) for Photonic Cross Connects" 
(PXC) http://search.ietf.org/internet-drafts/draft-sahay-ccamp-ntip-00.txt
<http://search.ietf.org/internet-drafts/draft-sahay-ccamp-ntip-00.txt>  
LMP-WDM proposes extensions to LMP to satisfy the OLI requirements, while 
NTIP proposes a new protocol.  We believe that the LMP approach is best 
because: 
1. [LMP-WDM] satisfies the OLI Requirements. 
2. The extensions to LMP are quite natural and fit within the spirit of the 
LMP protocol. 
3. LMP is a reasonably mature IETF protocol specification: 
     - It has been in the works for well over a year. 
     - It is an official working group document. 
4. It is better to have one protocol, than two for a given function 
(remember CR-LDP vs. RSVP-TE, and OSPF vs. IS-IS) 
5. Using the same protocol simplifies both implementation and management on 
devices, such as optical cross-connects and routers, that may need to use 
both simultaneously. 
While there is no doubt in my mind that NTIP could be made to satisfy the 
OLI requirements, It is my strong opinion that LMP should be the solution 
chosen by the CCAMP working group due to the reasons outlined above. 
Comments please! 
Regards, 
Andre