Hi I support adoption. This would allow for many ways for multicast to take advantage of flex-algo.
Regards, Stig On Thu, Aug 28, 2025 at 11:08 AM Peter Psenak <[email protected]> wrote: > > Hi Acee, > > On 28/08/2025 19:58, Acee Lindem wrote: > > > >> On Aug 28, 2025, at 1:26 PM, Peter Psenak <[email protected]> wrote: > >> > >> Hi Acee, > >> > >> On 22/08/2025 23:33, Acee Lindem wrote: > >>> Speaking as WG member: > >>> > >>> I think the requirements for this draft are "soft" given that the only > >>> example provided is the multicast distribution which has been satisfied > >>> for decades using a separate multicast topology. > >> MT does not give you the option to calculate paths based on various > >> metrics (e.g., delay, TE, bandwidth), use constraints, etc. That's what > >> flex-algo provides. > >> > >>> In short, the draft suffers from "ED" which in this case is "Exemplar > >>> Deficiency". I'm certainly not going to support adoption until I see a > >>> more compelling use case. > >> Flex-algo is strictly bound to a specific data-plane today - e.g., > >> SR-MPLS. SRv6, IP-Algo. The reason for that is that an algo specific > >> forwarding is needed and these data-planes are giving you that. > >> > >> The use case in this draft is to make flex-algo available for applications > >> that require algo specific paths without the need for any of these > >> data-planes. Such Flex-algo paths are not used for forwarding directly, > >> but rather by applications that create their own data-plane with the help > >> of the flex-algo paths. You can think of it as a distributed path > >> calculation service that provides you the loop free constraint based paths > >> on top of IGP topology for any app that needs it. > > What I’m saying is the forwarding plane doesn’t have to exist today for the > > multicast topology be used for multicast RPF. Please capture the reasoning > > as to why a separate data plane is required. Point out that while > > flex-algorithm can be topology specific, the SR-MPLS and IP flex algorithms > > are NOT topology specific. You guys understand this - just put it in the > > draft. > > sure. The new data plane is required, because in flex-algo routers must > advertise their participation in flex-algo and that participation is > data-plane specific. > > thanks, > Peter > > > > > Thanks, > > Acee > > > > > >> thanks, > >> Peter > >> > >> > >>> Thanks, > >>> Acee > >>> > >>> > >>>> On Aug 22, 2025, at 5:22 PM, Acee Lindem <[email protected]> wrote: > >>>> > >>>> LSR WG, > >>>> > >>>> The begins the LSR WG adoption call for "IGP Flex Soft Dataplane" - > >>>> draft-ginsberg-lsr-flex-soft-dataplane-01. Please express your support > >>>> or objection on this list prior to Saturday September 6, 2025. > >>>> > >>>> Thanks, > >>>> Acee > > > > _______________________________________________ > Lsr mailing list -- [email protected] > To unsubscribe send an email to [email protected] _______________________________________________ Lsr mailing list -- [email protected] To unsubscribe send an email to [email protected]
