I am trying to understand how this change could {help,hinder} the APNIC problem of route objects which reference differently maintained AS and Inetnum objects, and the added complication of AS being vested from RIPE and Inetnum from APNIC. 

I think stripping/changing the notify-on-ref mail might hinder this. It would be materially useful to preserve it, should exported IRR state be used at another site aggregating data, to contact the prime information manager.

I suspect RPSS/RPSAUTH issues are out of scope. Within one IRR/RPSL data set, I think the notify-on-ref thing would help the APNIC problem: it would make it easier for non-related maintainer to understand changes were being made in routing by address holders who want their AS to be in a route object, and participate.

-George



On Wed, Jun 11, 2014 at 5:13 AM, Sander Steffann <sander@steffann.nl> wrote:
Hi Job,

> I think some notification feature would be nice to have, but we need to
> figure out what and when we expect notifications.
>
> I propose we dub the attribute for nice alignment with existing
> attributes:
>
>    notify-on-ref: <email-address>      optional, multi-valued
>
> Questions:
>
>    - do you want a notification each time an object is updated and has
>      a reference to your object?

Strong no

>    - or do you only want notifications when a reference inititally is
>      added to an object? (spares you a daily mailbomb for daily updated
>      objects)

Yes

>    - do you want a notification when the reference is removed from an
>      object?

Yes

>    - In what classes do you want to set a notify-on-ref attribute? (I
>      think initially aut-num, as-set, rd-set)

Ack

>    - do we want the notify-on-ref email addresses to be set to
>      unread@ripe.net upon NRTM/ftp export?

No strong opinion on this one. I would say yes, unless someone comes up with a reason not to.

> Regarding authorisation, for me requiring authorisation to reference a
> given object is a bridge too far at this point in time. Quite some
> operators automatically generate an autnum, route-sets & as-sets on a
> daily basis to reject their policy, and I don't see an easy way to make
> this a painless adventure. Let's first do notifications and based on
> those experiences look further. ok?

Yes, that sounds reasonable. Needing authorisation to be allowed to put information in the policy sounds like a good way to discourage people from updating/using them altogether. Let's not make things more difficult unless we really need to.

Cheers!
Sander