On Mon, May 9, 2016 at 6:27 AM, Martin Bjorklund <[email protected]> wrote:

> Andy Bierman <[email protected]> wrote:
> > On Thu, May 5, 2016 at 7:44 AM, Balazs Lengyel <
> [email protected]>
> > wrote:
> >
> > > Hello,
> > >
> > > 1) Will mounted YANG modules be listed in the base ietf-yang-library?
> Even
> > > the mounted ones?
> > >
> >
> > Does the server providing the mounted modules claim conformance
> > to those models?
>
> No, which is why they are not listed in the base ietf-yang-library.
>
> > IMO YANG mount breaks the "YANG API contract" if even 1 tiny
> > property is ignored, due to mounting.
>
> Agree.
>
> > I asked for a conformance enum of 'none' or 'other' for this exact
> purpose
> > but
> > the WG did not think it was possible for a server to do anything
> > other than implement or import a YANG module.
> >
> > E.g, we are building a special server "library-mode" that just
> > provided\s the library and get-schema for every module.
> > They will all be listed as 'imported' (so that enum essentially
> > becomes 'other')
> >
> >
> > > 2) What does conformance to a YANG  module e.g. ietf-system mean now?
> Do we
> > > need to load the module at the top level, or is it enough to mount it
> > > “somewhere” in the containment tree?
> > >
> >
> >
> > It means the same if the server returns "implement"
>
> Yes.
>
> > Mounting the module violates the contract because the
> > module clearly defines /system as a top-level node.
>
> No, mounting doesn't break the contract, since the contract (base
> ietf-yang-library) doesn't mention these modules.
>
> > Any other location is non-compliant.
>
> I agree it would be non-compliant to list a mounted module in the base
> ietf-yang-library.  This is why mounted modules are not listed.
>
>

OK -- good.



> > > 3) If the answer to 2) is it can be mounted anywhere, which modules
> must be
> > > present at the top level? E.g. ietf-yang-library?
> > >
> >
> > IMO the YANG contract only supports the data placement explicit
> > in the module.
>
> Agreed.
>
> > > 4) If a new list entry is created can it introduce a completely new
> module
> > > never heard of before in the network element?
> > >
> >
> > ??
> >
> > this is fine for "anydata"
> >
> > > 5) IMHO allowing such dynamic introduction of new YANG modules into a
> > > network element, with flexible mountpoints, flexible set of modules at
> each
> > > mountpoint, a flexible set of deviations and features for each mounted
> > > module makes discovery actual available schema tree difficult. I think
> we
> > > are providing to much flexibility, over-complicating the issue. Some
> more
> > > static mounting solution that can be read from YANG modules instead of
> > > run-time data would be easier.
> > >
> >
> > mount is way over-engineered
>
> What do mean when you write "mount"?
>
> > but schema-defined mount-points don't really
> > help.
> > A nested "root" works and has a use-case.
>
> This is all the current schema mount does.
>
>

This is the only variant I support.

Renaming or relocating data (changing the ancestor path to the first root)
is way too complicated and not relevant to a programmatic interface.
This can be done in the presentation layer of the controller or application.




> > Everything else is extra
> > and an implementation choice without any real use-case.  YANG is already
> > complicated and slow. mount and sym-links makes it worse.
>
>
> /martin
>
>
>

Andy


>
> >
> > regards Balazs
> > >
> > >
> > >
> >
> > Andy
> >
> >
> > >
> > > --
> > > Balazs Lengyel                       Ericsson Hungary Ltd.
> > > Senior Specialist
> > > Mobile: +36-70-330-7909              email:
> [email protected]
> > >
> > >
> > > _______________________________________________
> > > netmod mailing list
> > > [email protected]
> > > https://www.ietf.org/mailman/listinfo/netmod
> > >
> > >
>
_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to