Re: [Sidrops] New Version Notification for draft-sriram-sidrops-drop-invalid-policy-00.txt

[Job Snijders <[e-mail]>]

Dear Jeff,

On Thu, Mar 15, 2018 at 07:36:12PM -0400, Jeffrey Haas wrote:
> On Tue, Mar 06, 2018 at 04:34:17PM +0000, Sriram, Kotikalapudi (Fed) wrote:
> > We have requested the chairs for time on the SIDROPS meeting agenda
> > to discuss this work:
> >  
> > https://tools.ietf.org/html/draft-sriram-sidrops-drop-invalid-policy-00 
> > 
> > The authors would appreciate comments/discussion on the list as
> > well.
> 
> I have a concern with one of the terms of your procedure.  From section 3:
> 
> : 4. If selected route is Valid/NotFound, then add the route to Loc-
> :    RIB; Else, if Invalid, then add the route to Loc-RIB only if
> :    there is no existing route in the Loc-RIB for a subsuming Less
> :    Specific prefix.
> 
> This sort of thing tries to get periodically added to BGP specs.  In
> general, BGP tries *VERY* hard to avoid any sort of cascading
> dependency between routes due to the ripple effects.  Enough so that
> the following is enshrined in the core spec:
> 
> RFC 4271, Section 9.1:
> 
> : The function that calculates the degree of preference for a given
> : route SHALL NOT use any of the following as its inputs: the existence
> : of other routes, the non-existence of other routes, or the path
> : attributes of other routes.
> 
> I appreciate what you're trying to do here.  However, my strong advice
> is DON'T.

I'd appreciate if you can elaborate more on your advice (perhaps from a
protocol architecture perspective). In quite some BGP implementations
the concept of "Conditional advertisements" is considered a feature,
which is seemingly at odds with RFC 4271 section 9.1.

Kind regards,

Job