[OSPF] Re: Comments on draft-berger-ospf-rfc2370bis-00.txt

[Acee Lindem <[e-mail]>]

 Hi Lou,
Lou Berger wrote:
> Acee,
>         See below.
>
> At 06:37 PM 11/20/2006, Acee Lindem wrote:
>
>> Hi Lou,
>> See inline.
>>
>>
>> Lou Berger wrote:
>>> Acee,
>>>         See responses in-line below.
>>>
>>> Should the corrected version go in as draft-berger or draft-ietf?
>>>
>>> Lou
>>>
>>> At 07:03 PM 11/13/2006, Acee Lindem wrote:
>>>
>>>> Hi Lou,
>>>>
>>>> My main comment is on section 5 (6 is your draft). The draft adds
>>>> mandatory reachability checking for AS scoped opaque LSAs.
>>>> If we add mandatory reachability checking, it should be for all
>>>> opaque types rather reading like the constraint is unique to AS scoped
>>>> opaque LSAs (type 11s).
>>>
>>> valid point.
>>>
>>>> I also think we should put the question as to whether the checking
>>>> should be madatory or relaxed a bit to allow an application to check
>>>> less frequently if the opaque data is stale.
>>>
>>> What benefit would this have?
>> This would allow applications that don't require the opaque data to
>> strictly track the SPF to discontinue using the information less 
>> frequently
>> than the SPF. This would be useful if OSPF and the application using the
>> opaque data were not tightly coupled. Here is what I put in the OSPFv3
>> update draft:
>>
>> 3.4.4.  Future LSA Validation
>>
>>   It is expected that new LSAs will be defined that will not be
>>   processed during the Shortest Path First (SPF) calculation as
>>   described in Section 3.8.  For example, OSPFv3 LSAs corresponding to
>>   information advertised in OSPFv2 using opaque LSAs [OPAQUE].  In
>>   general, the new information advertised in future LSAs should not be
>>   used unless the OSPFv3 router originating the LSA is reachable.
>>   However, depending on the application and the data advertised, this
>>   reachability validation MAY be done less frequently than every SPF
>>   calculation.
>>
>>   To facilitate inter-area reachability validation, any OSPFv3 router
>>   orginating AS scoped LSAs is considered an AS Boundary Router (ASBR).
>>
>> On the other hand, there is a school of thought that would consider 
>> this an
>> implementation detail. I, however, think it is better if it is 
>> explicitly stated.
>> If, for no other reason, to provide guidance to the people testing 
>> the protocols.
>
> Consistency between v2 and v3 is goodness where possible, so I'll add 
> a comment (to the end of 3.1) and reference.
Great.
>
>>>> Detailed comments:
>>>>
>>>> Section 2, third paragraph - "aligned" rather than "qaligned".
>>>
>>> yes.
>>>
>>>> Section 3, "Section 7" rather than "Section 7."
>>> okay.
>>>
>>>
>>>> Section 3.1 - Type 9  LSA - "keep" rather than "keepk".
>>>
>>> yes.  Also I added the following to the beginning of the section:
>>>  Section 13 of [OSPF] describes the OSPF flooding procedure.
>>>  Those procedures MUST be followed as defined except where
>>>  modified in this section.
>>>
>>>> I believe we
>>>>                    should discard a link-local LSA received from a 
>>>> neighbor not
>>>>                    on the interface (text similiar to type 11).
>>>
>>> okay, updated as follows:
>>>   o If the Opaque LSA is type 9 (the flooding scope is link-local)
>>>     and the interface that the LSA was received on is not the same
>>>     as the target interface (e.g., the interface associated with a
>>>     particular target neighbor), the Opaque LSA SHOULD be discarded
>>>     and not acknowledged, and MUST NOT be flooded out that interface
>>>     (or to that neighbor).  An implementation SHOULD keep track of
>>>     the IP interface associated with each Opaque LSA having a
>>>     link-local flooding scope.
>> I guess I think that if you discard an LSA, it is implied that you won't
>> reflood it.
>
> it does say "and MUST NOT be flooded".  I'm open to alternate wording.
In every other instance where an LSA is discarded in RFC 2328 we don't
explicitly state that we don't reflood them. In other words, why would
anybody get the misconcept we'd ever reflood anything that was discarded.
>
>
>
>>>> Section 3.1, Since the area ID is not in the LSA header, the bullet on
>>>>                   area flooding is confusing. It should say 
>>>> something to the
>>>>                   effect of only flooding type 10 LSAs out 
>>>> interfaces in the
>>>>                   LSA's associated area.
>>>
>>> okay, updated as follows:
>>>   o If the Opaque LSA is type 10 (the flooding scope is area-local)
>>>     and the area associated with Opaque LSA (as identified during
>>>     origination or from a received LSA's associated OSPF packet
>>>     header) is not the same as the area associated with the target
>>>     interface, the Opaque LSA MUST NOT be flooded out the interface.
>>>     An implementation SHOULD keep track of the OSPF area associated
>>>     with each Opaque LSA having an area-local flooding scope.
>> I think the action should be same as above. In any case, this can never
>> happen as the entire Link State Update packet would be discarded.
>
> okay, language is a hold-over from 2230.  will update to match above. 
> Same fore type 11.
Great.
>
>
>>>> I don't care if it said this in RFC 2370
>>>>                   and everyone knew what it implied.
>>>
>>> I don't know how to interpret the text otherwise...
>> RFC 2328 specifically state that packets whose Area ID cannot
>> be matched to the receiving interfaces are discarded. See section 8.2.
>
> I was agreeing with you...
Ok.
>
> [...]
>>>> 5.1 Inter-Area Considerations
>>>>
>>>> ......
>>>>
>>>>
>>>> Section 5.1
>>>>
>>>>   Type-9 opaque LSAs and type-10 opaque LSAs do not have this problem
>>>>   as a receiving router can detect an a loss of reachability 
>>>> through the intra-area
>>>>   SPF calculation.
>>>>
>>>> Section 5.1
>>>>
>>>>   To enable OSPF routers in remote areas to check availability of the
>>>>   originator of link-state type 11 opaque LSAs, the orignators of
>>>>    type-11 opaque LSAs are considered Autonomous System Border
>>>>    Routers (ASBRs) and will advertise themselves as such.
>>>
>>>
>>>
>>>>
>>>> Section 5.1 - Remove "It is important to note that this solution 
>>>> MUST NOT ..."
>>>> This is redundant.
>>>
>>> which part is redundant, just the sentence you are asking to be 
>>> removed?  I agree that it is redundant with 2328, but I thin 
>>> mentioning it is still useful.  Will rephrase to remove directive.
>> This part of the specification doesn't modify stub area behavior.
>> Here, we are talking about validating AS scoped opaque LSA.
>> I guess you could say, "Note that AS scoped opaque LSA
>> validation is not applicable to stub and NSSA areas since LSAs
>> with AS scope are not flooded into these areas types." However,
>> I don't see it as necessary.
>
> here's what I have now: "It is important to note that per
> [OSPF] this solution does not apply to OSPF stub areas or NSSAs as
> neither type-11 LSAs are flooded nor are type-4 LSAs originated into
> such areas."
>
> As you mentioned above, it is worth explicitly stating things sometimes.
Okay - I'll concede though I don't think it is necessary. Can we
call say "AS scoped opaque LSAs" rather than type-11s. Also, do
we have to add the detail about the ASBR-summary-LSAs (as they
are referred to RFC 2328).

>
>>>> Remove numbered items (1) and (2), these actions ARE NOT new to
>>>> opaque LSAs. Make (3) a separate paragraph rather than numbered
>>>> item.
>>>
>>> But inclusion of type-11 originate routers as ASBR is new.  Will 
>>> rephrase to make clear that existing ospf requirements apply.
>>>
>>> How about:
>>> The procedures related to inter-area opaque LSAs are as follows:
>>>
>>> (1) An OSPF router that is configured to originate AS-scope opaque
>>>     LSAs advertise themselves as ASBRs and MUST follow the related
>>>     requirements related to setting of the  Options field E-bit in
>>>     OSPF Hello packets and LSA headers as specified in [OSPF].
>>>
>>> (2) When ....
>> I don't think these points need to be restated even if they reference
>> RFC 2328. Opaque LSAs don't modify these conditions. Hence,
>> I don't see them to be required any more than stating the version field
>> the OSPF packet header should be set to 2.
>
> huh? opaque LSAs aren't present in 2328 and the specification of E-bit 
> setting/handling in 2328 is largely relative to AS-external-LSAs.
The E options bit must be set in hellos for any regular (not stub or 
NSSA) area
independent of any opaque support. Opaque LSAs don't change this at all.

Thanks,
Acee
>
>
> Thanks again for the comments,
> Lou
>

_______________________________________________
OSPF mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/ospf