Re: [netmod] 答复: FW: a question about ietf-hardware yang module

[Martin Bjorklund <[e-mail]>]

Juergen Schoenwaelder <[e-mail]> wrote:
> On Fri, Jun 28, 2019 at 09:52:20AM +0200, Martin Bjorklund wrote:
> > Juergen Schoenwaelder <[e-mail]> wrote:
> > > On Thu, Jun 27, 2019 at 10:04:50PM +0200, Martin Bjorklund wrote:
> > > > Juergen Schoenwaelder <[e-mail]> wrote:
> > > > > On Thu, Jun 27, 2019 at 09:52:56PM +0200, Martin Bjorklund wrote:
> > > > > > > Yes, good point, I think the phrase "by a different hardware
> > > > > > > component" should be removed. It seems last-change should change when
> > > > > > > I unplug a component and I plug it back shortly after, i.e., a
> > > > > > > component is replaced by itself. So we have:
> > > > > > > 
> > > > > > >         "The last time a new hardware component has been added to the
> > > > > > >          '/hardware/component' list, or a hardware component has been
> > > > > > >          removed from the '/hardware/component' list, or a hardware
> > > > > > >          component in the '/hardware/component' list has been
> > > > > > >          replaced."
> > > > > > 
> > > > > > I think that this is still not clear what it means that a component
> > > > > > has been replaced.  Do you mean "replaced by a different hardware
> > > > > > component"?
> > > > > > 
> > > > > > Otherwise (unplug then plug in the same component), the system either
> > > > > > detects the removal and thus updates last-change, or it doesn't detect
> > > > > > the quick removal/insertion, and then it can't do anything.  Thus, I
> > > > > > don't think this case needs special treatment, and the text could be
> > > > > > just:
> > > > > > 
> > > > > >          "The last time a new hardware component has been added to the
> > > > > >           '/hardware/component' list, or a hardware component has been
> > > > > >           removed from the '/hardware/component' list."
> > > > > >
> > > > > 
> > > > > The question is whether every implementor will figure out that if the
> > > > > component found in some slot x-y-z is different from what is expected
> > > > > to be in slot x-y-z, this must be seen as a remove + add combination.
> > > > > If we include 'replace', then it may be clearer that even in the case
> > > > > where what is in slot x-y-z has changed, the last-change must be
> > > > > updated. (That is, the list element with the same name still exists
> > > > > but it is different from what was there before with the same name.)
> > > > 
> > > > But then we're back to where we started - what exactly does "different
> > > > from what was there before" mean?  Presumably that some leaf's value
> > > > is different...?
> > > >
> > > 
> > > Go back some emails, I have removed the 'different' phrase.
> > 
> > Perhaps we can say:
> > 
> >        [...] or a hardware
> >        component in the '/hardware/component' list has been
> >        replaced with a new hardware component with the same
> >        name.
> >
> 
> Do we want last-change to be updated when a component gets unplugged
> and put back, i.e., a component got replaced with itself?
> 
> Can last-change be used to detect that something messed around with
> the components (or a component is connected via a bad line)? Or is
> last-change more strictly for the set of components has changed?
> 
> If I remove a component and I put it back slowly enough, I might get
> two last-change updates (the removal and the addition). If a client
> polls even more slowly, it will see that last-change has changed but
> it will find the same components. Hence, it seems reasonable to not
> exclude a replacement or a component with itself since it is a matter
> of the time resolution with which changes are detected and reported
> via the operational datastore.

Agreed.


/martin