Re: [netmod] working group adoption of draft-lhotka-netmod-yang-metadata-00

[Juergen Schoenwaelder <[e-mail]>]

On Thu, Jan 08, 2015 at 09:38:57AM +0100, Ladislav Lhotka wrote:
> Kent Watsen <[e-mail]> writes:
> 
> >>  - who believes a mechanism to define metadata annotations
> >>    should be standardized
> >
> >
> > Not yet.  
> >
> > First I think we need to clear the way for attributes to be used safely.
> > We used attributes in <edit-config> and also with-defaults, but in both
> > cases the client either accepted the contract of the protocol (base:1.1)
> > or explicitly opted into receiving the attributes (report-all-tagged).
> > Specifically, what happens to a legacy client if the attribute MUST be
> > semantically understood (e.g., the enabled/inactive) in order to correctly
> > process the data?   Even if we don't care about the client, we might
> > want
> 
> You are right but I think we have two separate issues here:
> 
> 1. Annotations are defined via an extension, and that by definition
>    means the client is free to ignore them.
> 
> 2. The existing sentiment of letting a (legacy) client pick only a
>    subset of modules from those advertised by the server means that the
>    client in any case cannot be expected to support an annotation that
>    is defined in a YANG module.
> 
> As for #1, it would help to make "annotation" a built-in statement, and
> that was actually one of the questions I asked earlier:

I believe we need to work with extensions, we can't revise the YANG
specifications each time an addition is needed.

My understanding is that this particular extension provides the
mechanics to define and encode metadata. It will be up to concrete
metadata definitions to figure out how to make things work such that
ignorant clients and servers will continue to function.

/js

-- 
Juergen Schoenwaelder           Jacobs University Bremen gGmbH
Phone: [phone]         Campus Ring 1, 28759 Bremen, Germany
Fax:   [phone]         <http://www.jacobs-university.de/>