On Tue, Nov 01, 2016 at 11:07:19AM -0400, Jeffrey Haas wrote:
> On Sun, Oct 30, 2016 at 05:10:01PM +0100, Job Snijders wrote:
> > I'd like to propose a different strategy and describe some background.
> > Both of you mention per-peer flags or new per-route state flags, which
> > in essence are updates to rfc4271, this feels like too heavy of a
> > hammer.
> > 
> > I'd prefer to make the document compatible with existing deployments of
> > software which is already compliant with the spirit of this document.
> > 
> > OLD:
> >     Software MUST mark any routes from an EBGP peer as 'invalid' in the
> >     Adj-RIB-In, if no import policy was configured.
> > 
> > NEW:
> >     Software MUST consider any routes from an EBGP peer as invalid, if
> >     no import policy was configured.
> > 
> > NEW2:
> >     "Software MUST discard any routes from an EBGP peer, if no import
> >     policy was configured."
> 
> I rather object to NEW2 and, if included, withdraw any support of this
> draft.
> 
> A fundamental issue with this behavior is that it dumps routes that
> would have to be recovered via expensive refresh.

Am I correct to understand that the word 'discard' has very specific
meaning in this context? Does "discard" mean "forbidden to store in
memory"?

Sriram, I believe you were the first to raise a concern about the "OLD"
text, do you have a preference for "NEW2" or strongly object to "NEW"?
If the latter, can you help bridge the gap between intended meaning
implementation consequences?

Or perhaps: NEW3: "Software MUST ignore any routes from an EBGP peer, if
no import policy was configured." ?

> FWIW, I still haven't evaluated our own software under the prior
> version of the draft, but think it's a generally good idea.

I look forward to your evaluation!

Kind regards,

Job

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

Reply via email to