[quagga-dev 5466] Re: Unnumbered interface support in ospf

J.J. Krabbendam jkrabbendam at aimvalley.nl
Wed Jun 4 08:59:34 BST 2008


Joakim Tjernlund wrote:
>   
>>>>> LSA-generation.
>>>>> The normal situation for point-to-point interfaces is that a router
>>>>> LSA will be generated containing two link descriptions entries (see
>>>>> 12.4.1.1 of RFC 2328): a type 1 (point-to-point) and a type 3 (stub
>>>>> network). It appeared that a CISCO box using its unnumbered feature
>>>>> only uses a type 1 link description.
>>>>>       
>>>>>           
>>>> Seems like current zebra always add a Type 3, stub network. What do you
>>>> think?
>>>>     
>>>>         
>> Depends on what you want. If you are REALLY ptp unnumbered, the type 3
>> LSA does not make sense since neither option 1 nor option 2 are
>> applicable (RFC 2328: 12.4.1.1). If you have numbered ptp (or re-used
>> interfaces) then probably you should fill in data according to option
>> 1.
>>     
>
> Even if you are using REALLY unnumbered ptp, I think OSPF should send a
> type 3 option 1 LSA anyway as any packets from such an interface needs
> to be assigned a real IP address. Does something break if you do send a
> type 3 LSA to the CISCO box? 
>
>  Jocke
>
>   
I agree that packets that go out of the unnumbered interface need and ip 
address. The kernel just fetches an address
of one of the other interfaces. So that address should better be 
advertised through OSPF, otherwise one has connectivity problems.
Do you agree that neither option 1 not option 2 that are mentioned in 
rfc2328:12.4.1.1 are applicable?
I do not have a CISCO box here. A co-worker is also teaching at a school 
and he had the possibility to do a test at the school for me.



-- 
This message has been scanned for viruses and is believed to be clean

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quagga.net/pipermail/quagga-dev/attachments/20080604/a25f2a80/attachment-0001.html>


More information about the Quagga-dev mailing list