Hi, Stewart, On Tue, Dec 4, 2018 at 6:07 AM Stewart Bryant <[email protected]> wrote:
> > > On 30/11/2018 19:23, Spencer Dawkins at IETF wrote: > > This is Mirja's comment, but ... > > On Fri, Nov 30, 2018 at 10:12 AM Mirja Kühlewind <[email protected]> > wrote: > >> Mirja Kühlewind has entered the following ballot position for >> draft-ietf-opsawg-ipfix-bgp-community-11: No Objection >> >> When responding, please keep the subject line intact and reply to all >> email addresses included in the To and CC lines. (Feel free to cut this >> introductory paragraph, however.) >> >> >> Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html >> for more information about IESG DISCUSS and COMMENT positions. >> >> >> The document, along with other ballot positions, can be found here: >> https://datatracker.ietf.org/doc/draft-ietf-opsawg-ipfix-bgp-community/ >> >> >> >> ---------------------------------------------------------------------- >> COMMENT: >> ---------------------------------------------------------------------- >> >> One comment on section 1: >> "For example, they can shift some flows >> from congested links to low utilized links through an SDN controller >> or PCE [RFC4655]." >> I'm not aware that ipfix information is commonly used for dynamic traffic >> adaptation and I'm not sure that is recommendable. C > > > I'm agreeing with Mirja here. > > We've spent a LOT of time not recommending dynamic traffic adaptation. > Probably half my responsibility as AD for ALTO was repeating "you can't > react based on changes to that attribute without taking chances on > oscillation" like it was a mystical incantation, and I wasn't the first AD > to have that conversation repeatedly. > > > Yes, I understand the ARPA net had exactly that problem at one stage and > had to be converted from using a delay based metric to a fixed metric. > > > I would be happy to hear that all those problems are solved, but I haven't > heard that yet. Do the right thing, of course. > > Even "can shift some flows from persistently congested links to > underutilized links" would cause me less heartburn. > > > There is no such thing as permanent in network paths! > > Like many control problems the first order solution is to damp with a > suitably long time constant, but infinity, i.e. permanent, is not a > satisfactory choice either. > Yeah, that's where I was headed (stated more coherently). Saying "I should do something, because the network path is STILL congested" is safer than "I should do something because the network path is congested", so now we're talking about suitable definitions of "STILL". I was shooting for that with "persistent", and agree that "permanent" path characteristics is a happy idea we aren't likely to see in practice. Do the right thing, of course ;-) Spencer > - Stewart > > > Spencer > > > > >
_______________________________________________ OPSAWG mailing list [email protected] https://www.ietf.org/mailman/listinfo/opsawg
