[Note that I'm responding to things in thread-order.  I2RS was very busy for
a few days while I had to pay attention to other things, so my responses may
appear to jump around a bit.]

Sue,

On Fri, Sep 26, 2014 at 12:22:25PM -0400, Susan Hares wrote:
> 
> Just to confirm three facts before I put them in i2rs drafts (based on
> Juergen's email
> 
> - based on Juergen's email:   I2RS is "config true empheral".  

This is a point of discussion.

A critical thing to realize from the netmod interim is that it was simply a
first step to trying to define what "ephemeral" even *means* in the context
of existing yang and netconf/restconf.  The "option 4" solution was the
proposal that worked its way through discussion points with the yang
experts, but still needs discussion within I2RS and then related followup
with netconf if we decide it's the way to go.

I like the idea, but as discussion is pointing out, there's a lot of details
to reconcile.

The impact of this upon writing data models for I2RS would appear to be at
this point:
- They're just "config true" (per correction from Juergen), but simply done
  in the ephemeral datastore.  How you document that they're just in the
  ephemeral data store is unclear at this point.
- They're otherwise just normal models.
- But if they're intended to be i2rs ephemeral extensions on local config
  models (e.g. protocol WG data models), we have some dependency work to
  reconcile with those groups.

> If i2rs has additional configuration for BGP related to your I2RS peers for
> reporting routes: 
> 
>  
> 
> .         BGP based config + I2RS additional config + AS Config
> 
>  
> 
> Can I2RS have additional configuration values specific to I2RS?  For
> example, information regarding how often to report the routes for a
> particular pear?  

Correct.  

We have roughly four scenarios with regard to modeling impact:
A. Completely disjoint models for I2RS that use no state from other protocol
   modules and no dependencies, but interact with underlying state of that
   protocol.  E.g. BGP configuration in an I2RS model that doesn't extend a IDR
   Yang module nor have constraints in such a module.
B. A disjoint model for I2RS that has constraints against other modules.
   E.g. the I2RS BGP module may have a "must" statement against a IDR yang
   module's "autonomous system" configuration.
C. A hybrid I2RS module that augments an underlying protocol module.  This
   is the scenario you're suggesting above.  The critical detail is since it's
   an augmentation it is suggesting there exists such a protocol module.
D. There is a unified module developed in some working group that may simply
   choose to store some of its state in an ephemeral fashion.

D is the newest option enabled by the "option 4" discussion from the netmod
interim.

-- Jeff

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

Reply via email to