Re: [netmod] schema-mount pre09 branch

[Ladislav Lhotka <[e-mail]>]

On Wed, 2018-02-07 at 13:27 +0100, Martin Bjorklund wrote:
> Hi,
> 
> It seems we're now just reiterating what has previously been discussed
> *a lot*.  IMO, highest prio is to resolve any issues related to YLbis.

I've always missed a technical merit in those discussions. It has always been
like "I prefer that solution", "This is not how our implementation works", "It
is good enough for LNE/NI" etc. But maybe I missed something so please point me
to discussions that demonstrate why the adopted solution is better than what I
am proposing. I am arguing that it is exactly the other way around. The inline
and use-schema cases are different concepts and mixing them together makes the
whole thing needlessly complex to describe and understand.

> If we also need other clarifications to make the document easier to
> understand, that's fine.  But I don't think we should fundamentally
> change the solution that the WG agreed upon.

I don't agree that we are *strongly* changing the solution. Simplifications and
clarifications are IMO badly needed.

Lada  

> 
> 
> /martin
> 
> 
> 
> Ladislav Lhotka <[e-mail]> wrote:
> > On Wed, 2018-02-07 at 11:41 +0000, Robert Wilton wrote:
> > > 
> > > On 07/02/2018 10:29, Ladislav Lhotka wrote:
> > > > On Wed, 2018-02-07 at 11:14 +0100, Martin Bjorklund wrote:
> > > > > Juergen Schoenwaelder <[e-mail]> wrote:
> > > > > > On Tue, Feb 06, 2018 at 03:25:52PM +0000, Robert Wilton wrote:
> > > > > > > I think that the term "external" could also be confusing, since I
> > > > > > > think
> > > > > > > that
> > > > > > > sort of implies peer mount like semantics.
> > > > > > 
> > > > > > The "inline" mount concept seems to subsume peer mounts. From the
> > > > > > model perspective, is there a difference whether the mounted data is
> > > > > > local or remote (and what does local/remove mean for a VM)?
> > > > > >   
> > > > > > > I would suggest the term "dynamic" instead of "inline " but that
> 
> could
> > > > > > > easily be confused with dynamic datastores.
> > > > > > 
> > > > > > Yes, I think this is not a good word either.
> > > > > > 
> > > > > > > Perhaps rather than "inline" another choice could be
> 
> "discoverable",
> > > > > > > i.e.
> > > > > > > the schema is not known, and is dynamically discoverable inline at
> 
> the
> > > > > > > mount
> > > > > > > point.
> > > > > > > Equally, rather than "use-schema", perhaps a better choice would
> 
> be
> > > > > > > "known",
> > > > > > > i.e. the schema is already known, and made available as part of
> 
> YANG
> > > > > > > library.
> > > > > > 
> > > > > > Perhaps integrated schema vs. mounted schema.
> > > > > 
> > > > > I like the term "integrated" better than "use-schema".  But both cases
> > > > > are mounted, so we need another term than "mounted" for "inline".
> > > > > "segregated" doesn't sound quite right ;-)
> > > > 
> > > > I would prefer to use the term "mount" only for the inline case and find
> > > > something else for the use-schema case. The term "mount" evokes that
> 
> some
> > > > *instance* data being added, which is what happens in the "inline" case
> 
> but
> > > > not
> > > > for "use-schema".
> > > 
> > > Perhaps the "use-schema" case really is a type of "schema mount", where 
> > > as the "inline" case is a type of "mount".
> > 
> > This may be quite confusing. My suggestion for "use-schema" is "external
> > augment" - the mount point as a *schema node* plays a very similar role to
> 
> the
> > target node of an augment.
> > 
> > > 
> > > Perhaps they could/should have entirely separate YANG models to describe 
> > > them.  Possibly in the "use-schema" case could refer to grafting a 
> > > schema into a parent schema rather than mounting it.
> > 
> > I proposed this previously. The inline case could in fact be considerably
> > simplified because the extension statement is all that's needed - no state
> 
> data.
> > In other words, the "mount-point" extension would immediately indicate the
> > inline mount.  
> > 
> > In order to distinguish the use-schema case (or whatever we call it) we have
> > then two options:
> > 
> > 1. use a different YANG extension for labelling mount points of this type
> > 
> > 2. use schema node identifiers as in augments (i.e. no extension at all).
> > 
> > Thanks, Lada
> > 
> > > 
> > > Thanks,
> > > Rob
> > > 
> > > 
> > > > 
> > > > Lada
> > > > 
> > > > > 
> > > > > /martin
> > > > > 
> > > > > > > Whether it would be right to change these at this time, I've no
> 
> idea
> > > > > > > ...
> > > > > > 
> > > > > > Yep.
> > > > > > 
> > > > > > /js
> > > > > > 
> > > > > > -- 
> > > > > > Juergen Schoenwaelder           Jacobs University Bremen gGmbH
> > > > > > Phone: [phone]         Campus Ring 1 | 28759 Bremen |
> 
> Germany
> > > > > > Fax:   [phone]         <https://www.jacobs-university.de/>
> > > > > > 
> > > > > 
> > > > > _______________________________________________
> > > > > netmod mailing list
> > > > > [e-mail]
> > > > > https://www.ietf.org/mailman/listinfo/netmod
> > > 
> > > 
> > 
> > -- 
> > Ladislav Lhotka
> > Head, CZ.NIC Labs
> > PGP Key ID: 0xB8F92B08A9F76C67
> > 
> > _______________________________________________
> > netmod mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/netmod
> > 
-- 
Ladislav Lhotka
Head, CZ.NIC Labs
PGP Key ID: 0xB8F92B08A9F76C67