Hi Acee,

On 12/30/13 18:27 , Acee Lindem wrote:
Hi Peter,

On Dec 29, 2013, at 6:22 AM, Peter Psenak wrote:

Hi Acee,

some clarification is required in terms of EL-bit setting in various compatibility cases. 
I assume EL-bit is set only in "Normal" mode.

can you please clarify the EL-bit setting in various compatibility modes?


Regarding the AreaExtendedLSASupport, you defined only two modes for it, 
MixedMode is missing, not sure why. Also, if we keep both global and per area 
compatibility support, some intersection analysis between the two is required - 
e.g. do we support 'MixedMode' at global level with 'Normal' at area level - 
these two contradict each other in terms of adjacency formation with legacy 
routers.

We could support all three values at the area level as well. My intention is 
that the area setting would take precedence over the global level for 
area/link-local scoped LSAs in the area and area adjacency formation. I see 
what you mean about a contradiction. Let me see if I can clear it up.

I see.



AreaExtendedLSASupport

    * Global - Support for Extended LSAs is controlled by the value of 
ExtendedLSASupport.
    * Normal - Extended LSAs will be originated for Area and Link-local scoped 
LSAs within the area. Adjacencies are not formed with Routers not supporting 
the Extended LSAs. The handling of AS External LSA is controlled by the value 
of ExtendedLSASupport.
    * MixedMode -  Both extended and non-extended LSAs will be originated for 
Area and Link-local scoped LSAs within the area. Adjacencies are not formed 
with Routers not supporting the Extended LSAs. The handling of AS External LSA 
is controlled by the value of ExtendedLSASupport.

should not MixeMode above allow adjacency to be formed with routers not supporting the Extended LSAs? MixedMode at global level allows that, which is what we want.

thanks,
Peter



I'm afraid that keeping the compatibility support at two different levels is 
going to create a lot of confusion.

I don't want to add more confusion. I was merely looking for a way to migrate 
one area at a time to the extended LSA format.

Thanks,
Acee




thanks,
Peter


On 12/28/13 23:37 , Acee Lindem wrote:
Given that the authors really don't want to leave a legacy of complexity
with the backward compatibility cases. I think we are going to reduce
the compatibility cases to:

ExtendedLSASupport
       This is an enumeration type indicating the extent to which the
       OSPFv3 instance supports the TLV format described herein for
       Extended LSAs.  The valid value for the enumeration are:

       *  None - Non-extended LSAs will not be originated or used in the
          SPF calculation.

       *  Normal - Extended LSAs will be originated and adjacencies will
          not be formed with OSPFv3 routers not supporting this
          specification.

       *  MixedMode - Both extended and non-extended LSAs will be
          originated.  OSPFv3 adjacencies will be formed with OSPFv3
          routers not supporting this specification.  The non-extended
          LSAs are used for the SPF computation.


One thing that occurred to me is that it might be useful to migrate a single 
area. In this case, one would allow the following modes:

AreaExtendedLSASupport
       This is an enumeration type indicating the extent to which the
       OSPFv3 area supports the TLV format described herein for
       Extended LSAs.  The valid value for the enumeration are:

       *  None - Non-extended LSAs will not be originated or used in the
          SPF calculation.

      * Normal – Area and link-local scoped Extended LSAs will be originated 
and adjacencies will not be formed with OSPFv3 routers in the area not 
supporting this Extended LSAs. AS scoped LSAs will be originated as 
non-extended LSAs.


Thoughts?


Thanks,

Acee







_______________________________________________
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