Re: [Netconf] RGW YANG push -18 LC comments [part 1]

[Martin Bjorklund <[e-mail]>]

Hi,

I have one comment on this thread.

Robert Wilton <rwilton=[e-mail]> wrote:
> On 06/09/2018 20:35, Alexander Clemm wrote:

[...]

> >         (13) Rather than "No Sync on start", it might be better to refer to
> >         this
> >
> >         as  "sync on start", which could be defined as a boolean with default
> >
> >         true [I've not checked the YANG model]
> >
> >     <ALEX> it's not defined as a boolean but as a leaf of type empty. 
> >     (So, the "default" is in fact true, unless  no-sync-on-start is
> >     present).  I don't see a compelling reason to change this and suggest
> >     to keep this as-is.
> >
> >     </ALEX>
> >
> > OK.  I think that in YANG, particularly considering NMDA, it is
> > cleaner to model this as a boolean with default true, particularly if
> > you do a get-data request on the operational state, since it more
> > clearly and explicitly indicates whether sync-on-start is enabled.
> >
> > With the current model, on a get-data request against operation, the
> > server returns "no-synch-on-start", it is obvious that it won't
> > perform a sync-on-start, however, the reverse if the leaf isn't
> > present isn't quite so clear.  If it doesn't return
> > "no-synch-on-start" then that either means the device will perform
> > sync-on-start, or perhaps the node the server isn't returning any
> > information, or perhaps it has been deviated.  So, in summary, I think
> > that flags that turn something off, are better modeled with default
> > true, rather than as empty leafs.  Flags that turn something on can be
> > modeled equally well either way.
> > <ALEX-2>In terms of instantiation, I find empty leaves more elegant,
> > particularly if there is a default value.  I think <no-sync-on-start/>
> > is more elegant than <no-sync-on-start>True</no-sync-on-start>.
> > Would you be satisfied if we turn the default around, i.e. change
> > “no-sync-on-start” to “sync-on-start”, but keeping it an empty?
> >  </ALEX-2>
> RW: Yes, changing it to "sync-on-start" keeping the type as empty is
> fine.

But this would change the default behaviour that the WG has accepted.
I don't think this is a good change.  I agree with Robert's initial
comment that this should be modeled as a leaf 'sync-on-start' of type
boolean, with default true.  I prefer the better model before a more
elegant encoding.


/martin