Joel,

Please see inline.

On 4/18/14, 5:19 PM, "Joel M. Halpern" <[email protected]> wrote:

>Actually Jan, you made a leap there I did not follow.
>You assert that "the controller programmatic API iis a superset of a
>device programmatic API.”
>This does not seem to follow.  On the one hand, I would hope that a
>controller is exposing abstractions, which are unrelated to the device
>API.  

Yes. I may have stated slightly differently - I called these abstractions
“controller’s own services”.

>And on the other hand, there may well be many things the device API can
>do that the controller does not
>need.  

Again, we are in agreement, with one possible twist: an application may
want to talk to the device not directly, but through the controller (I
think a while back we all agreed that this was a valid use case). I called
it “controller proxying the device”.

The point I was trying to make that Yang, as the IDL in the controller, is
suitable both for defining controller’s own service APIs and for proxying
of device APIs. I spoke from experience: we defined a bunch of controller
services and proxied a bunch of device APIs (netconf itself, PCEP,
OpenFlow) with yang models. It worked really well for all APIs that we
managed to throw at it :-).

To conclude: if yang works well in the controller environment, which is a
more complex API environment than a device, it will surely work well at
the device level. That’s all that I wanted to say.

>The I2RS charter makes it clear that I2RS is not intended to replace
>other configuration mechanisms.
>
>And the basic premise of I2RS is that there are requirements for the
>work that were not addressed properly by the existing configuration
>protocols.  Otherwise the WG would not even need to discuss protocol
>modifications.  So the fact that NetConf / YANG works for device
>configuration does not seem to be evidence that it works for what I2RS
>wants to do.
>
>I would actually guess that it is reasonably close,

Again, we seem to be in agreement - I also said that the majority of use
cases that I can think of can be covered with Netconf/Restconf/Yang today.
I did not say we can cover everything. But, through the OpenDaylight
experience I found that Netconf/Yang could cover a lot more than what I
originally anticipated (which was actually quite surprising to me).


>and I will live with whatever conclusion the WG reaches about protocol
>and modeling.  But I would not claim that the existing deployment is
>relevant evidence for this problem space.

I actually think it is - what’s wrong with working code? ;-)

>
>Yours,
>Joel


Thanks,
Jan

>
>On 4/18/14, 7:59 PM, Jan Medved (jmedved) wrote:
>> 
>> 
>> On 4/18/14, 2:12 PM, "Jamal Hadi Salim" <[email protected]> wrote:
>> 
>>> On Fri, Apr 18, 2014 at 4:59 PM, Jan Medved (jmedved)
>>><[email protected]>
>>> wrote:
>>>> Cisco is also implementing Netconf - it¹s available on XR today, and
>>>>it
>>>> will be available on other platforms as well.
>>>>
>>>> For OpenDaylight, we chose Yang as the IDL to describe internal and
>>>> external APIs in the controller and so far it has served its purpose
>>>> really well.
>>>>
>>>
>>> This has nothing  to do with meeting requirements for I2RS.
>> 
>> Sure it does. I2RS is ultimately about a set of programmatic APIs to the
>> routing system. The controller programmatic API is a superset of a
>>device
>> programmatic API, since the controller must be able to proxy the device
>> and provide programmatic APIs to its own services. Now, if an IDL has
>> proven to be useful and/or meet the requirements in a environment which
>>is
>> a superset of the target environment, then by definition it will be
>>useful
>> and/or meet the requirements in the target environment.
>> 
>>>
>>>> Also, as Tom pointed out, Netconf and Restconf have also been
>>>> implemented
>>>> in ODL. I¹d like to stress that we not only have multiple
>>>> Netconf/Restconf
>>>> implementations from multiple vendors (Brocade, Juniper, Cisco, just
>>>>to
>>>> mention those on this thread), but have multiple open source
>>>> implementations as well. In addition to ODL, there is libnetconfd and
>>>>a
>>>> few others.  ODL / libnetconfd interop testing is done in the ODL
>>>> regression test suite.
>>>
>>> Again - I am not hearing requirements against I2RS.
>>> Have you implemented I2RS with netconf/restconf/yang?
>> 
>> Nobody has implemented I2RS, because the protocol has not been defined
>> yet. But I was deeply involved with Netconf/Restconf/Yang
>>implementations
>> on both IOS-XR and on OpenDaylight. In any case, Netconf/Restconf/Yang
>> gives us today maybe 80-90% of what we need from I2RS for our use cases.
>> 
>>>
>>>> Now, since there are multiple independent open
>>>> source implementations, we¹ve got a good ecosystem for implementation
>>>>of
>>>> new Netconf/Restconf/Yang features that may be required to meet i2rs¹s
>>>> needs (if the need to evolve the protocols/language arises).
>>>>
>>>
>>> I dont think this sounds rational at all. HTTP has a bigger ecosystem
>>> than you.
>>> Why not build around that and then refactor as needed?
>> 
>> Bingo! How about Restconf with JSON encoding? ;-)
>> 
>>> Why is it so hard to get requirements around here?  What is the point
>>> to a standard again?
>> 
>> The I2RS requirements have been gathered, and Netconf and Yang have been
>> analyzed against the requirements by Andy and Dean at the last IETF. I
>> find no fault in their analysis, so IMHO Netconf / Yang (with possibly
>> small modifications) will meet the technical requirements of I2RS.
>> 
>> What I and others have been trying to say is that the rest of the world
>>is
>> moving towards Netconf/Restconf/Yang. Therefore, (besides it being a
>> good-enough technical choice), it is the pragmatic choice. If I2RS wants
>> to be relevant to operators, vendors, and the open source community, it
>> needs to go where all these entities want to be.
>> 
>>>
>>> cheers,
>>> Jamal
>> 
>> 
>> Thanks,
>> Jan
>> 
>>>
>>>> In short, +1 for Netconf/Restconf for the i2rs protocol and for Yang
>>>>as
>>>> the modeling language.
>>>>
>>>>
>>>>
>>>> Thanks,
>>>> Jan
>>>>
>>>>
>>>> On 4/18/14, 6:17 AM, "Dean Bogdanovic" <[email protected]> wrote:
>>>>
>>>>> Jamal,
>>>>>
>>>>> Here are two criteria to be considered:
>>>>> 1. technical
>>>>> 2. commercial/business
>>>>>
>>>>> We can discuss pros and cons for both, but have to state that from
>>>>> business perspective for Juniper going with RESTCONF/YANG make more
>>>>> sense. We already built the Junos model in YANG and
>>>>> have or are in process of building needed tools. Same goes for
>>>>>RESTCONF.
>>>>> We have NETCONF implemented in Junos and are working on RESTCONF
>>>>> implementation.
>>>>> Many carriers adopted or are adopting NETCONF/YANG, are looking into
>>>>> RESTCONF as well, so this looks like a low hanging fruit from
>>>>>business
>>>>> perspective.
>>>>>
>>>>> Looking at technical aspect, unless there is a very compelling reason
>>>>> (and there might be, but I'm not aware of it), don't see reason to
>>>>> switch
>>>> >from RESTCONF and YANG.
>>>>> We can find out down the line that RESTCONF/YANG was the wrong way to
>>>>> go,
>>>>> but that can be always changed. From my perspective it looks right
>>>>> today.
>>>>>
>>>>> Just to be clear, I vote for RESTCONF/YANG adoption for i2rs.
>>>>>
>>>>> Dean
>>>>>
>>>>> On Apr 18, 2014, at 7:27 AM, Jamal Hadi Salim <[email protected]>
>>>>>wrote:
>>>>>
>>>>>> Ok, since nobody is saying anything i'll bite.
>>>>>> How would you like for this discussion to proceed?
>>>>>>
>>>>>> On Fri, Apr 11, 2014 at 1:50 PM, Edward Crabbe <[email protected]>
>>>>>>wrote:
>>>>>>> Dear I2RSers,
>>>>>>>
>>>>>>>   At the last I2RS WG meeting there was a great deal of
>>>>>>>conversation
>>>>>>> regarding selection of both modeling language and underlying
>>>>>>> transport
>>>>>>> protocol.  Consensus at the time was to make use of Yang and
>>>>>>>(NetConf
>>>>>>> or
>>>>>>> RestConf) (unclear).
>>>>>>>
>>>>>>
>>>>>> And i believe the view, as correctly presented by you, is for folks
>>>>>>to
>>>>>> go back
>>>>>> and make educated decisions by actually getting knowledgeable about
>>>>>> the different views presented. "Consensus" that you described above
>>>>>> to me looked like  a pageant popularity contest not based on
>>>>>>anything
>>>>>> technical ("who likes contestant in the blue shirt? please cheer for
>>>>>> them").
>>>>>>
>>>>>> In my opinion i dont think the requirements are clear.
>>>>>>
>>>>>> Will that get the crickets stop chirping?
>>>>>>
>>>>>> cheers,
>>>>>> jamal
>>>>>>
>>>>>>>   Before coming to a final consensus, we'd like to give people
>>>>>>> adequate
>>>>>>> time
>>>>>>> to review source material, marshall arguments and discuss on the
>>>>>>> mailing
>>>>>>> list.  To this end, we're asking that interested parties do just
>>>>>>>this
>>>>>>> over
>>>>>>> the course of the next ~two weeks. Following that period, on 4/28,
>>>>>>> we'll be
>>>>>>> initiating a consensus call that will last an additional two weeks,
>>>>>>> with the
>>>>>>> aim of converging modeling language / protocol by Friday, 5/9.
>>>>>>>
>>>>>>> The consensus call should also generate proposals for any material
>>>>>>> changes
>>>>>>> required to the underlying protocols.  These proposals in turn will
>>>>>>> form the
>>>>>>> basis for a later draft including gap analysis and said changes.
>>>>>>> Those
>>>>>>> strongly in favor of one protocol over another should be prepared
>>>>>>>to
>>>>>>> contribute to this analysis.
>>>>>>>
>>>>>>>
>>>>>>> best,
>>>>>>>
>>>>>>>   -ed
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> i2rs mailing list
>>>>>>> [email protected]
>>>>>>> https://www.ietf.org/mailman/listinfo/i2rs
>>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> i2rs mailing list
>>>>>> [email protected]
>>>>>> https://www.ietf.org/mailman/listinfo/i2rs
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> i2rs mailing list
>>>>> [email protected]
>>>>> https://www.ietf.org/mailman/listinfo/i2rs
>>>>
>> 
>> _______________________________________________
>> i2rs mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/i2rs
>> 
>
>_______________________________________________
>i2rs mailing list
>[email protected]
>https://www.ietf.org/mailman/listinfo/i2rs

_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to