On Wed, Apr 20, 2022 at 06:19:09PM -0700, Andy Bierman wrote:
> 
> If the argument is that a zone index is always allowed (even if the
> usage is limited) then why does the ip-address-no-zone typedef exist
> at all?  There are no YANG guidelines for picking between them.
>

A simple rule of thumb is that the no-zone type is adequate if the
zone in which an address is valid is available from the context.

- Example: If a data model configures a link-local address on an
  interface, then the interface provides the context and hence the
  zone can be omitted.

- Example: If a data model refers to an IP endpoint and there is no
  further context information (e.g., you configure the IP address of a
  DNS server), then the data model should allow for a zone index to be
  present.

I believe this is _architecturally_ the right thing to do.

The other dimension is how we deal with implementations that choose to
not support zoned addresses where they architecturally make sense. One
option is to let such implementations declare deviations. Another is
to let such implementations reject zoned addresses for leafs where
zoned addresses are rightfully supported by the data model but not by
the implementation. What we should not encourage is to have
implementations that accept zones addresses but then silent strip away
the zone since this leads to unpredictable behaviour.

/js

-- 
Jürgen Schönwälder              Jacobs University Bremen gGmbH
Phone: +49 421 200 3587         Campus Ring 1 | 28759 Bremen | Germany
Fax:   +49 421 200 3103         <https://www.jacobs-university.de/>

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

Reply via email to