Juergen Schoenwaelder <[email protected]> writes:

> On Thu, Feb 04, 2016 at 11:46:47AM -0500, [email protected] wrote:
>>
>> Lou Berger <[email protected]> writes:
>>
>> > Juergen,
>> >
>> > see below
>> >
>> > On 2/4/2016 11:12 AM, Juergen Schoenwaelder wrote:
>> >> On Thu, Feb 04, 2016 at 09:53:11AM -0500, Lou Berger wrote:
>> >>> Juergen,
>> >>>
>> >>> see below.
>> >>>
>> >>> On 2/4/2016 9:12 AM, Juergen Schoenwaelder wrote:
>> >>>> On Thu, Feb 04, 2016 at 08:02:26AM -0500, Lou Berger wrote:
>> >>>>>> If more granular mounts are needed, then we should IMHO _not_ bundle
>> >>>>>> this with the notion of YANG submodules. Perhaps you meant submodules
>> >>>>>> in a more generic way, but then perhaps:
>> >>>>>>
>> >>>>>> s/of submodules/of parts of modules/
>> >>>>> yes.
>> >>>> OK - so I will read submodules as 'parts of modules'.
>> >>>>
>> >>>>>> Reading the other text again, I am not sure what is meant by the
>> >>>>>> phrase "incorporation of the data model defined by one top-level
>> >>>>>> module". What exactly is a 'top-level module' (and does it matter,
>> >>>>> interfaces.
>> >>>> An example does not define the term.
>> >>> 100% agree - at least in drafts.
>> >>>
>> >>>> Please define 'top-level module'
>> >>> any module that defines a top-level node, or if you prefer a module that
>> >>> defines nodes at the XPath root node.
>> >>>
>> >>>> so we can actually understand what we are talking about.
>> >>> If you don't like either formation, I suspect at this point you know
>> >>> what I mean, so please propose alternate language that works for you and
>> >>> I'll confirm that/if it works for me.
>> >> Cool. So if I mount ietf-interfaces (a top-level module) into some
>> >> other place, what happens to all the data models that augment
>> >> ietf-interfaces with interface specific objects, like the ietf-ip
>> >> module (a non-top-level module)?
>> >
>> > they are allowed/supported in the same way that they would be today,
>> > just with the modifications/rewrite of their base models.
>> >
>> >>
>> >> I fail to see why this distinction between top-level modules and
>> >> non-top-level modules is useful.
>> >
>> > I'm describing a single use case, which only requires top-level support,
>> > not all desirable capabilities or a solution.
>>
>> Hi Juergen,
>>
>> The top-level modules would carry the mounts of the non-top-level
>> modules that are attached within them. I don't see this making sense any
>> other way. The idea here is to support a parent device being able to
>> mount a contained child device's modules and modify them external to the
>> child devices normal management mechanism (e.g., not having to log into
>> the child's netconf server).
>>
>> Given that, what's the use of talking about mounting "inner" modules?
>> You have to mount the "outer" (or containing) module to be able to
>> access them, and by mounting the containing module you gain access to the
>> contained modules.
>>
>> At least that's how I picture this working.
>>
>
> But even then, why do we need new terms? If you mount a path, then you
> mount a path in a datastore. Modules that do not define paths obviously
> then can't be mounted. How about this:
>
> module top-level-module {
>   container top;
> }
>
> module non-top-level-module {
>   import "top-level-module" { prefix top; }
>
>   augment "/top:top/" {
>     container second;
>   }
> }
>
> Can I mount /top/second? Why would a solution be simpler if I make
> this impossible?


So your saying the parent can the second without the first on the
/top/second path? If so then it should be the same with the "child" (or mount
point) as well. The intent is not to try and add restrictions here.

I think maybe the whole top-level thing is unintentionally confusing.

Thanks,
Chris.

>
> /js

Attachment: signature.asc
Description: PGP signature

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

Reply via email to