Re: [i2rs] network-type: container vs. identity?

["Alexander Clemm (alex)" <[e-mail]>]

Hi Xian,

the way it is structured with containers and all will allow you to indicate the various network types that apply to a network.  If there is a hierarchy of network types, you would have a hierarchy of containers.  In your case  network-types --> ietf-te-topology--> otn-topology.  Network will indicate each of the network types that it instantiates.  

You could model a hierarchy of network types also through a hierarchy of identities, as you indicate in your example.  Sure, in your example this is a viable alternative, but it is not clear what you would gain through that.  For one, in the case of "otn-topology", the fact that the same topology is also an ietf-te-topology is rather implicit, not explicit.  Per Robert's point, if for some reason you have multiple traits and want to "compose" a network to encompass multiple types, this is straightforward to do with the current pattern, not so straightforward with identities.   

--- Alex


-----Original Message-----
From: i2rs [mailto:[e-mail]] On Behalf Of Zhangxian (Xian)
Sent: Tuesday, July 12, 2016 7:49 AM
To: Robert Varga <[e-mail]>; [e-mail]
Cc: [e-mail]
Subject: Re: [i2rs] network-type: container vs. identity?

Hi, Robert, 

   Thank you for your clarification. Just make sure I get your point, let me try with a (real) example. 

   At the moment, we have the following relationship among different topology types ( using the diagram from https://tools.ietf.org/html/draft-ietf-i2rs-yang-l3-topology-02 as base):

            +-----------------------------+
           |  +-----------------------+  |
           |  |    ietf-network |  |
           |  +----------^------------+  |
           |         |            |
           |  +-----------------------+  |
           |  | ietf-network-topology |  |-------< [ietf-te-topology]------<[ otn-topology] ( as an example)
           |  +----------+------------+  |
           +-------------^---------------+
                         |
                         |
             +-----------^-------------+
             | l3-unicast-igp-topology |
             +----+---------------+----+
                  ^               ^
                  |               |
                  |               |
         +--------^-----+      +-----^---------+
         | ospf-topology|    | isis-topology |
         +--------------+       +---------------+

So, this shows support of more than two levels of relationships and many branches. Is that what you mean by multiple traits and composition? Although I do not see an example using this identity for this, but I wonder if the following way of using identity is supported and can be used to cater your need? 

Identity network-types {
  Description "base type for network types"; }

Identity type-topology {
 Base "network-types";
}

Identity type-l3-unicast-igp {
 Base " type-topology ";
}

Identity type-ospf {
 Base "type-l3-unicast-igp";
}

I am cooking the codes up with my limited understanding of YANG, so I might be wrong. If so, please do let me know. 

Cheers,
Xian

-----Original Message-----
From: Robert Varga [mailto:[e-mail]]
Sent: 2016年7月11日 18:54
To: Zhangxian (Xian); [e-mail]
Cc: [e-mail]
Subject: Re: [i2rs] network-type: container vs. identity?

On 07/08/2016 05:47 AM, Zhangxian (Xian) wrote:
> Hi, Authors,
> 
>  
> 
>    while using this model as a base to augment for technology-specific 
> topologies, I wonder why the leaf network-types is a container, 
> instead of being as a identity?
> 
>  
> 
> I remember I asked Alex offline before, but the response I got was 
> that the identity was also under consideration at that time. Given the 
> latest version (June 2016 version) still use container, I wonder if 
> the authors can explain why the alternative is discarded? Thank you.

The idea is to be able to have multiple traits, for example for composition.

Bye,
Robert


_______________________________________________
i2rs mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/i2rs