Re: [GROW] Genart last call review of draft-ietf-grow-bgp-gshut-11

[David Farmer <farmer@xxxxxxx>]

On Tue, Oct 10, 2017 at 7:33 AM, <bruno.decraene@xxxxxxxxxx> wrote:

> From: Job Snijders [mailto:job@xxxxxxx]
 > Sent: Tuesday, October 10, 2017 2:00 PM
>
 > On Tue, Oct 10, 2017 at 11:41:32AM +0000, bruno.decraene@xxxxxxxxxx wrote:
 > >  > Any attribute (origin, as_path, aggregator) anywhere can be overloaded
 > >  > to mean something only significant to the local network. I think the
 > >  > document is simpler without this and see no point in mentioning this. I
 > >  > propose:
 > >  >
 > >  > OLD:
 > >  >     The LOCAL_PREF value must be lower than the one of the alternate
 > >  >     path. 0 being the lowest value, it can be used in all cases, except
 > >  >     if it already has a special meaning within the AS.
 > >  > NEW:
 > >  >     The LOCAL_PREF value SHOULD be lower than any of the alternative
 > >  >     paths. It is RECOMMEND to use 0, the lowest LOCAL_PREF value.
 > >
 > > What is really needed is "The LOCAL_PREF value SHOULD be lower than
 > > the one of the alternative path." Looks reasonable to extend it to
 > > your proposition " The LOCAL_PREF value SHOULD be lower than any of
 > > the alternative paths." So I'm changing for this.
 > >
 > > Now the value is truly local to an AS, and I'm not sure to see the
 > > technical reason to RECOMMEND (SHOULD) a specific value. MAY seems
 > > more appropriate to me. Hence I'm proposing to keep "Zero being the
 > > lowest value, it MAY be used whichever LOCAL_PREF values are used by
 > > the AS."
 >
 > So the total of the new text is as following?
 >
 >     "The LOCAL_PREF value SHOULD be lower than any of the alternative
 >     paths.  Zero being the lowest value, it MAY be used whichever
 >     LOCAL_PREF values are used by the AS."

Yes, that is correct.

 > I am not sure about the second sentence, it seems hard to read.

I'm open to reformulation.

 > I see value in just recommending a value for people moving between ASNs
 > (debugging other organisation's networks via BGP looking glasses) to
 > recognise as a highly undesirable path.

I agree.

> Reading RFC 2119 the
 > 'RECOMMENDED' seems appropiate, "use 0 unless you have a reason not to".

I'm fine with that part, but the subsequent RFC 2119 text "but the full implications must be understood and carefully weighed before choosing a different course." seems too strong for me, as there is just no issue with an AS choosing a different value.


 > This is a GROW document and I believe clear-cut guidance will benefit
 > all.

OK. What about using lower case "recommended" ?
Proposed NEW: Zero is the lowest value and MAY be used whichever LOCAL_PREF values are used by the AS, hence the use of LOCAL_PREF 0 is recommended.

(possibly adding "for consistency between ASes and implementations" )

Thanks again for your comments.
Kind regards,
--Bruno

I would prefer a normative RECOMMENDED, the rest of the sentence in RFC2119, just means you should explain the constraints on the alternatives. How about something like this; 

"The LOCAL_PREF value SHOULD be lower than any of the alternative paths.  A LOCAL_PREF value of Zero is RECOMMENDED, however any LOCAL_PREF value lower than all other LOCAL_PREF values used within an AS is an acceptable alternative. The LOCAL_PREF value used, Zero or otherwise, SHOULD NOT also have another use or meaning within the AS."

--

===============================================
David Farmer               Email:farmer@xxxxxxx
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota  
2218 University Ave SE        Phone: 612-626-0815
Minneapolis, MN 55414-3029   Cell: 612-812-9952
===============================================