Re: [netconf] [yang-doctors] Yangdoctors last call review of draft-ietf-netconf-subscribed-notifications-21

[Andy Bierman <[e-mail]>]

On Mon, Jan 21, 2019 at 1:49 AM Martin Bjorklund <[e-mail]> wrote:

> Andy Bierman <[e-mail]> wrote:
> > On Wed, Jan 16, 2019 at 10:35 AM Eric Voit (evoit) <[e-mail]>
> wrote:
> >
> > > Hi Andy,
> > >
> > > Thanks for the review.  Some thoughts....
> > >
> > > > From: Andy Bierman, January 14, 2019 8:08 PM
>
> [...]
>
> > > > I7) leaf stream-xpath-filter: [multiple uses]
> > > >
> > > >            The expression is evaluated in the following XPath
> context:
> > > >
> > > >              o   The set of namespace declarations is the set of
> prefix
> > > >                  and namespace pairs for all YANG modules implemented
> > > >                  by the server, where the prefix is the YANG module
> > > >                  name and the namespace is as defined by the
> > > >                  'namespace' statement in the YANG module.
> > > >
> > > >   -- This prefix processing is not done anywhere else in NETCONF
> > > >      or RESTCONF.  IMO a bad precedent.  Only the XML prefixes
> > > >      should be required for processing of XML encoding.  YANG
> > > >      module prefixes are not required to be unique, unlike
> > > >      the prefix mappings in XML
> > >
> > > This text was proposed by Martin as a result of the "xpath expressions
> in
> > > JSON" thread last October in NETMOD.
> > >
> > > I am happy to incorporate whatever text is appropriate.  I was hoping
> that
> > > the suggested text was sufficient for now.  Kent has already
> incorporated
> > > this as an issue for yang-next
> > > https://github.com/netmod-wg/yang-next/issues/55
> > > So hopefully there is no final precedent being claimed.
> > >
> > >
> > I do not agree that this YANG module should define a new way to encode
> XPath
> > into XML instance documents.
>
> This mechanism was first used in RFC 8040, section 4.8.4, in order to
> have a transport-agnostic encoding of XPath expressions (same for XML
> and JSON).
>
> There was long debate on the ML for this issue, and this was the
> outcome that the WG decided.
>
>
OK -- no need to change anything then.
I do not agree that "xmlns" attributes pose an implementation challenge for
clients
but servers already need custom XPath to process YANG so this just adds to
that list.



> > This will require significant changes to server
> > implementations.  YANG module prefixes are not even required to be unique
> > so the set of prefixes used by the server in XML instance documents may
> be
> > different,
> > since it must be unique.
>
> YANG module prefixes are not used here.  The module *name* is used as
> the identifier prefix in XPath expressions:
>
>   /ietf-interfaces:interfaces/ietf-interfaces:interface/...
>
>
>
> /martin
>


Andy