Hi Paul,

On May 4, 2010, at 1:12 PM, Paul Wells wrote:

> Hi Acee,
> 
> Acee Lindem wrote:
>> Speaking as WG member:
>> 
>> I am not opposed to this capability. However, I'm not sure if it justifies 
>> taking two of the remaining router bits. The Router-Information LSA (RFC 
>> 4970) is really the place for advertising capabilities.
> 
> We understand this concern and are in the process of drafting a 
> new revision that moves the "S" bit to the RI opaque LSA. Note, 
> though, that the definition of the Router Informational 
> Capabilities TLV in RFC 4970 explicitly says "...the informational 
> capabilities advertised have no impact on the OSPF protocol's 
> operation..." so we're defining a new "operational capabilities" 
> TLV to hold this and any future "important" bits.

I remember we made the RFC 4970 statement due to debate during IESG review 
regarding backward compatibility for existing capabilities. Adding a new TLV 
for new capabilities that do impact protocol operation sounds like a great 
idea. 


> 
>> It would be ideal if this could be done without an implicit backward 
>> compatibility mechanism. Doing an election just to see if all routers in the 
>> area are able to recognize the R-bit is complexity that would be nice to  
>> avoid. The draft recommends both setting the R-bit and advertising any 
>> transit links with 0xffff cost. Hence, It would be an odd topology where the 
>> differences in R-bit capability could cause a routing loop. Unfortunately, 
>> it is possible to draw one if you there is a valid path to a destination D, 
>> and that destination is accessible via multiple routers with varying R-bit 
>> capabilities :^(. 
> 
> It would be nice, but I don't see any alternative that can be 
> guaranteed to be correct for all topologies.

We discussed this over a beer and the only thing we could come up with was 
putting the configuration and responsibility in the hands of the system 
administrator (since IGPs are under the control of a single administrative 
domain). However, this would be a departure from what we've done heretofore in 
the OSPF WG. 

Thanks,
Acee 



> 
> Thanks,
> Paul
> 
>> 
>> Thanks,
>> Acee 
>> 
>> 
>> On Apr 16, 2010, at 4:44 PM, Acee Lindem wrote:
>> 
>>> At the OSPF WG meeting in Anaheim, the subject draft was presented and 
>>> there seemed to be some interest in such a capability (similar to the 
>>> overload bit in ISIS). Please indicate whether or not you support making 
>>> this draft a WG document. 
>>> 
>>> http://www.ietf.org/id/draft-pillay-esnault-ospf-rbit-00.txt
>>> 
>>> Thanks,
>>> Acee - (as OSPF WG Co-chair) 
>>> _______________________________________________
>>> OSPF mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/ospf
>> 
>> _______________________________________________
>> OSPF mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/ospf

_______________________________________________
OSPF mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ospf

Reply via email to