Thanks. I think stating these things more clearly would help a lot. I also agree with Eric's point that explaining the *other* advantages of GAAO (beyond just efficiency) would help justify why this document defines an additional mechanism to assign prefixes beyond DHCPv6 PD.
On Wed, Jul 29, 2026 at 4:24 PM Luigi IANNONE <luigi.iannone= [email protected]> wrote: > Hi Lorenzo, > > > > > > *[email protected] <[email protected]>* > > Paris Research Center > > Huawei Technologies France S.A.S.U. > > > > *From:* Lorenzo Colitti <[email protected]> > *Sent:* Friday, July 24, 2026 12:15 PM > *To:* [email protected] > *Cc:* [email protected] > *Subject:* [6lo] DHCPv6 vs GAAO option > > > > draft-ietf-6lo-nd-gaao authors, > > > > Looking at https://datatracker.ietf.org/doc/html/draft-ietf-6lo-nd-gaao-11 > again, I still think that a better choice would be to reuse DHCPv6 instead > of adding the new GAAO option. > > > > Appendix A.3 shows that in terms of bytes and messages, DHCPv6 with Rapid > Commit is essentially as efficient as GAAO. The remaining inefficiency > mentioned in the draft is that currently, DHCPv6 SOLICITs are sent to > multicast addresses. > > *[LI] Correct.* > > This is easy to fix: define a well-known anycast address for the DHCPv6 > server to use in 6lo deployments (as Esko mentioned at the mic). That will > be exactly as efficient as the GAAO option if the server is on the > first-hop router. > > *[LI] Yes, could be a solution.* > > But it could also optionally (or in the future) support scenarios where > the server is not in the first-hop router. There is already an M bit in the > RA that I think could be used to signal support for this. > > *[LI] Possibly.* > > > > This seems better than the GAAO option because it does not require > defining a new protocol and provides additional flexibility. If I'm missing > something, or if we do not take this approach for some other reason, the > draft should explain why. > > *[LI] GAAO is an option in 6LoWPAN ND, not a protocol by itself. As stated > during the meeting, if you run a 6LoWPAN ND stack, having an additional > option is ‘cheaper” than running an additional service like DHCP. Agreed, > this needs to be stated more clearly. * > > *Note additionally that the document suggest the use of GAAO _only_ in > 6LoWPAN_ deployments and _only_ if addresses are assigned algorithmically. > More generic assignment policies in 6LoWPAN do need DHCPv6. Agreed that > this last sentence is not in the document, but we will add it, along with > the comments you provided above.* > > > > *Thanks* > > > > *Luigi * > > > > Cheers, > > Lorenzo >
_______________________________________________ 6lo mailing list -- [email protected] To unsubscribe send an email to [email protected]
