Andy Bierman <[email protected]> wrote:
> On Thu, Nov 5, 2015 at 5:12 PM, Martin Bjorklund <[email protected]> wrote:
> 
> > Andy Bierman <[email protected]> wrote:
> > > On Thu, Nov 5, 2015 at 4:50 PM, Martin Bjorklund <[email protected]> wrote:
> > >
> > > > Hi,
> > > >
> > > > Andy Bierman <[email protected]> wrote:
> > > > > Hi,
> > > > >
> > > > > I asked about this text on page 156:
> > > > >
> > > > >
> > > > >    In statements that have any data definition statements as
> > > > >    substatements, those data definition substatements MUST NOT be
> > > > >    reordered.
> > > > >
> > > > > I asked why this new rule was added.
> > > >
> > > > It is not a new rule, it is there in RFC 6020 as well.
> > > >
> > > >
> > > Then the word "reordered" needs to be clarified
> > >
> > >
> > >
> > > > In rpc/action input/output the nodes are encoded in XML in the order
> > > > they are defined.  So this rule is there to protect clients that might
> > > > depend on that order.
> > > >
> > >
> > >
> > > But this just applies to the advertised version of the RPC.
> > > The client should use the schema the server supports,
> > > and the order is dictated by that version.
> >
> > In general the update rules are there in order to protect old
> > clients.  If we say that the client needs to understand the revision
> > that the server implements, then we can remove all update rules.
> >
> > > > > I do not see why it is needed, or how it can be enforced.
> > > > >
> > > > > If I add a new object to grouping A, then every place there
> > > > > is a "uses A", a new data node will be inserted inline.
> > > >
> > > > Newly inserted nodes does not change the order between the old nodes.
> > > >
> > > >
> > >
> > > This is one interpretation of the term "reorder".
> > > This interpretation is OK -- the relative order cannot be changed
> > > but overall, any subtree can have nodes added in between
> > > existing objects (via direct edit or uses).
> >
> > Ok, how about:
> >
> > OLD:
> >
> >    In statements that have any data definition statements as
> >    substatements, those data definition substatements MUST NOT be
> >    reordered.
> >
> > NEW:
> >
> >    In statements that have any data definition statements as
> >    substatements, the relative order between old data definition
> >    substatements MUST NOT be changed.
> >
> > > > We could relax the rule and say that it only applies to input/output
> > > > and groupings (since they might be in input/output.
> > > >
> > >
> > >
> > > OK, except the identification of objects does not depend on order.
> > > This is kind of a CLR.
> >
> > Do you think we should relax the rule?
> >
> >
> yes.
> 
> One could argue that it might take operational experience to know
> that a certain RPC input order is more advantageous than the current order.
> IMO the places YANG is brittle mostly involve the statements where decisions
> have to be right the first try because the cost of starting over is too
> high,
> and YANG does not allow a new version to fix the problem.
> 
> 
> We won't change the order of <edit-config> because it is thought to be
> correct.
> If <target> was after <config> maybe we would want to change it, but YANG
> rules
> say that is forbidden.

Right.  In that case we'd have to use a new name for the operation.
Maybe it's ok.


/martin

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

Reply via email to