Re: [i2rs] I-D Action: draft-ietf-i2rs-yang-network-topo-13.txt

[Robert Wilton <[e-mail]>]

Hi Alex,

If you need to represent learned topologies before NMDA compliant 
implementations are available then you need the extra -state module 
(i.e. a copy of the NMDA compatible I2RS topology module, but with name 
appended with -state and all nodes set as config false).  This could be 
generated via tooling, put into github, or added in an appendix to the 
draft.

Without this, then the existing I2RS topology module can only be used to 
represent configured topologies on non NMDA compliant implementations 
(specifically any implementations that don't expose the operational 
state datastore).

For NMDA compliant implementations the network topology module in draft 
-13 works well.

Thanks,
Rob


On 26/06/2017 18:52, Alexander Clemm wrote:
>
> Hi Rob,
>
> Inline <ALEX>, below
>
> Thanks
>
> --- Alex
>
> ---------- Forwarded message ----------
> From: "*Robert Wilton*" <[e-mail] <mailto:[e-mail]>>
> Date: Mon, Jun 26, 2017 at 1:53 AM -0700
> Subject: Re: [i2rs] I-D Action: draft-ietf-i2rs-yang-network-topo-13.txt
> To: "Alexander Clemm" <[e-mail] <mailto:[e-mail]>>, 
> <[e-mail] <mailto:[e-mail]>>, "'Nitin Bahadur'" 
> <[e-mail] <mailto:[e-mail]>>, "'Russ 
> White'" <[e-mail] <mailto:[e-mail]>>, "'Xufeng Liu'" 
> <[e-mail] <mailto:[e-mail]>>, 
> <[e-mail] <mailto:[e-mail]>>, "'Jan Medved 
> (jmedved)'" <[e-mail] <mailto:[e-mail]>>, 
> <[e-mail] <mailto:[e-mail]>>, "'Susan 
> Hares'" <[e-mail] <mailto:[e-mail]>>, "Kent Watsen" 
> <[e-mail] <mailto:[e-mail]>>, "Martin Bjorklund" 
> <[e-mail] <mailto:[e-mail]>>
>
> Hi Juergen,
> On 24/06/2017 14:17, Juergen Schoenwaelder wrote:
> > On Thu, Jun 22, 2017 at 11:44:00AM +0100, Robert Wilton wrote:
> >> Do you think that it would be useful if the draft also included the extra
> >> transient "-state" modules in an appendix (e.g. as per
> >> draft-dsdt-nmda-guidelines-01 section 2)?
> >>
> >> Specifically, I'm thinking to help make the topology module fully usable by
> >> modules that augment it (e.g. by the TE modules if/when they adopt the NMDA
> >> conventions), until NMDA implementations before widely available.
> >>
> > Rob,
> >
> > the less we have of those transient "-state" trees, the better it is.
> > For LMAP (in auth48) we did not do this. These extra "-state" trees
> > should ideally only be used in very rare cases, I think existing code
> > already works with a single tree (at least this is what I understood
> > from the OpenDaylight discussions).
> I completely agree with you in general, but for the topology module I
> think that the -state tree is required to represent topologies that
> exist but have not been configured (e.g. perhaps those learned from a
> dynamic routing protocol).
> Also copying Kent and Martin, since they were very both very involved in
> the discussions on the I2RS alias discussing the structure of the I2RS
> network topology module.
> My interpretation is from Xufeng was it is needed for the TE YANG
> modules, but if it turns out that it is not actually needed, then that
> is also good with me ;-)
> <ALEX>
> The need to represent topologies that are learned is certainly there.  
> It is not exclusive to TE, and I would be surprised if TE YANG modules 
> have an extra need for a separate state tree.  Probably the best 
> person to comment here is Xufeng, but it sounds to me, also per 
> Juergen’s comments, that an extra state tree will _/not/_ be needed.
> </ALEX>
> Thanks,
> Rob
> >
> > /js
> >