>>> Right, but how about multiple NUMBERED PtoP interfaces with the >>> same local IP address? >> I'd argue that if your configuration supports this, all but one of >> the interfaces is unnumbered. >
Joakim, IP address must be unique and assigned to one numbered IP interface only. This has nothing to do with OSPF, it is requirement of a basic IP addressing paradigm and covered in relevant IPv4 architecture documents (no, I don't have a reference ready). This 'limitation' is the reason why unnumbered interfaces were invented in the first place. If "multiple NUMBERED PtoP interfaces with the same local IP address" worked well then nobody would bother to have unnumbered interfaces. Note that numbered interface by definition implies the interface has mask <32 bits and that remote end has IP address from the same IP subnet. If implementation allows configuration of IP address with mask of 32 bits or no mask at all then this in fact is just implementation of IP unnumbered interface where 'master' interface which owns the IP address is not explicitly shown. Conceptually it would be IP unnumbered nonetheless and should be treated by OSPF accordingly. Anton Joakim Tjernlund wrote: > Acee Lindem <[email protected]> wrote on 14/08/2009 16:23:15: >> On Aug 14, 2009, at 3:14 AM, Joakim Tjernlund wrote: >> >>> Acee Lindem <[email protected]> wrote on 14/08/2009 00:13:48: >>>>>> Is there something in the OSPF spec which prevents it from operate >>>>>> on multiple numbered PtoP interfaces with the same local IP >>>>>> address? >>>>>> >>>>>> I have that working, but there are arguments that this is >>>>>> allowed by >>>>>> OSPF, i.e, ethier use unnumbered links or unique local IP >>>>>> addresses. >>>>>> >>>>>> Jocke >>>>> Got some comments on this but nothing conclusive, I really want >>>>> to know. Acee, what do you think? >>>> Since there is only a type 1 (router link) in the Router-LSA for each >>>> unnumbered link, it makes no difference how many use the same IP >>>> address. >>> Right, but how about multiple NUMBERED PtoP interfaces with the >>> same local IP address? >> I'd argue that if your configuration supports this, all but one of >> the interfaces is unnumbered. > > hmm, please clarify. I do mean treating each and every PtP > as a numbered PtP link. Is there something that prohibits > this in the spec? Of course not all routers will > be able to handle this but in that case it would be a > limitation in the impl. > > Is it allowed to have numbered on one side of the link > and unnumbered on the other side? > >> >>> Then there will be multiple identical >>> type 3 too. >> Even though it is non-standard, I don't see any problems. I believe >> Dave alluded to the fact that some OSPF implementations may have ABR >> SPF optimizations which can have problems when the same address is >> advertised in multiple addresses - I know fixed some problems here in >> one implementation. > > Yeah, when one only uses the router LSAs to identify the interface > you are in trouble. > >> Acee >> >>> Jocke >>> >> > > _______________________________________________ > OSPF mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/ospf _______________________________________________ OSPF mailing list [email protected] https://www.ietf.org/mailman/listinfo/ospf
