Re: [netmod] Inventory YANG model (entity-MIB)

["Dongjie (Jimmy)" <[e-mail]>]

Hi Andy,

Thanks for your feedback. Please see my replies inline:

> -----Original Message-----
> From: Andy Bierman [mailto:[e-mail]]
> Sent: Tuesday, March 10, 2015 1:24 AM
> To: Juergen Schoenwaelder; Jeffrey Haas; Benoit Claise; NETMOD Working
> Group; [e-mail];
> [e-mail]
> Subject: Re: [netmod] Inventory YANG model (entity-MIB)
> 
> On Sun, Mar 8, 2015 at 1:29 PM, Juergen Schoenwaelder
> <[e-mail]> wrote:
> > On Sun, Mar 08, 2015 at 03:08:37PM -0400, Jeffrey Haas wrote:
> >> [speaking largely as a contributor and not i2rs-chair]
> >>
> >> On Fri, Mar 06, 2015 at 12:05:36PM +0100, Juergen Schoenwaelder wrote:
> >> > My first question is (perhaps not surprising) whether inventory
> >> > falls into the I2RS charter, I2RS = interface to the routing
> >> > system. That said, RFC 6643 gives you a read-only translation.
> >> > There are not many read-write objects in the ENTITY-MIB so perhaps
> >> > this is good enough for now. I guess it would help what I2RS needs
> >> > to know in order to make the interface to the routing system work.
> >>
> >> As discussed later in thread, having this work owned by netmod is
> >> fine with I2RS.
> >>
> >> The structure of the generic topology draft is leading to interesting
> >> questions about how information present in that model can link to
> >> higher and lower layers.  As seen in the I2RS presentations at the
> >> most recent interim, this eventually leads to questions like
> >> "inventory".  I also raised the question about tunnels which will lead to
> other models as well.
> >>
> >> The fact that the IETF already has good models for some of these
> >> things, the ENTITY-MIB as an example, is good.  However, simple
> >> conversion of MIBs to yang modules are likely to result in models
> >> that are good MIB replacements, but might not be well tailored to
> >> re-use elsewhere.  Since much of the motivation is for re-use, re-use
> >> is the discussion I would suggest is most important.
> >>
> >> As an example, SNMP tables probably should be converted into yang
> groupings.
> >> There's also the matter of indexing of such conversions may not be
> >> fully appropriate for such re-use and alternate indexes may be helpful.
> >
> > Nobody is insisting on an SNMP translation but whatever is done better
> > allows export via the ENTITY-MIB as well - things need to be aligned
> > at least. I note that many of the things you mention are not part of
> > draft-dong-i2rs-network-inventory-00.txt either.
> >
> 
> I start to worry about I2RS boiling the ocean when I see inventory data models
> in that WG. I question the value of converting every OID to a YANG identity, or
> making gratuitous name changes to every leaf.

The inventory model was motivated by some discussion in the I2RS interim, and we are open to move this work to Netmod if people are more interested to do this work here.
As Jeff said, simple conversion of MIB to Yang is not exactly our goal. Hopefully we can make good use of Yang to build a base inventory model, which is extensible with augmentation to build more specific models.

> Perhaps an SMI to YANG translation of the ENTITY-MIB will be a good enough
> starting point. I prefer to see IETF YANG work focus on configuration, rather
> than reinvent monitoring solutions that have been deployed for many years.

To my understanding, SMI to Yang translation gives you read-only models, while with Yang we can do both read and write. I'd agree that the inventory yang model should not limit its scope for monitoring, and we need to discuss what configurations should be considered as part of the inventory model. We'd appreciate your contributions on both aspects. Thanks.

Best regards,
Jie

> >> > PS: I personally would have preferred if generic topology and perhaps
> >> >     inventory would have split off into a short-lived targeted WG
> >> >     instead of doing all of this in I2RS but it seems leadership has
> >> >     already decided that I2RS is the home for all of this.
> >>
> >> [speaking as one of the i2rs-chairs]
> >>
> >> Topology was one of the use cases that was in-charter and we're not
> >> looking to significantly expand our charter work.  But one thing
> >> we've noted over the life of the group is that there has been a lot
> >> of push for micro-WG to be formed for various yang efforts.  I
> >> personally believe we're better off with some owning WG taking on the
> >> task but handling it within the context of a design team.  The overhead of a
> full WG is likely not merited.
> >>
> >> Which WG the design team reports to is, of course, up for discussion.
> >>
> >
> > What was in the original I2RS charter was routing topology export
> > (read-only). Going from there to a generic topology model (which
> > likely should not be limited to config false) is quite a step. I
> > personally find it sub-optimal to do the generic model in WG A and
> > several extensions of the generic model in WG B plus some more
> > extensions in WG C, most likely crossing IETF areas. The coordination
> > resulting from this will most likely slow things down.
> >
> > /js
> >
> 
> Andy
> 
> > --
> > Juergen Schoenwaelder           Jacobs University Bremen gGmbH
> > Phone: [phone]         Campus Ring 1 | 28759 Bremen |
> Germany
> > Fax:   [phone]         <http://www.jacobs-university.de/>
> >
> > _______________________________________________
> > netmod mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/netmod