Re: [Lsr] Comments on draft-wang-lsr-prefix-unreachable-annoucement

[Peter Psenak <[e-mail]>]

Zhibo,

On 28/07/2022 22:07, Huzhibo wrote:
> Peter
> 
>> -----Original Message-----
>> From: Lsr [mailto:[e-mail]] On Behalf Of Peter Psenak
>> Sent: Friday, July 29, 2022 8:33 AM
>> To: Aijun Wang <[e-mail]>; Acee Lindem (acee)
>> <[e-mail]>
>> Cc: Ketan Talaulikar <[e-mail]>; Les Ginsberg (ginsberg)
>> <ginsberg=[e-mail]>;
>> [e-mail]; lsr <[e-mail]>
>> Subject: Re: [Lsr] Comments on
>> draft-wang-lsr-prefix-unreachable-annoucement
>>
>> Aijun,
>>
>> On 28/07/2022 19:55, Aijun Wang wrote:
>>> Hi, Acee:
>>>
>>> Thanks for your comments, but most of them are indefensible,
>>> especially the conclusion.
>>> As you have also noticed, UPA mechanism doesn’t consider the network
>>> partition scenarios, doesn’t consider how to control the number of
>>> advertisement of unreachable messages, doesn’t provide the explicit
>>> notification of unreachable statement(as also pointed out Ketan, Bruno
>>> etc.), then how you hurry to get the conclusion that UPA is superior to PUA?
>>
>> IETF documents are not deployment guides, nor design or implementation
>> documents, not the source of education for the other vendors.
>>
>> IETF documents are there to specify the bare minimum to achieve
>> interoperability.
>>
>> In other words, the fact that you put more content in your document, does not
>> make it any better. Contrary, the less you need to do to achieve the
>> interoperability, the better it is.
> 
> [HZB] More content you mentioned of the documents are addresses comments on the promotion of this document.
> It is an essential part of a complete solution.
> RFC 5305 define that LSInfinity metric is used for other purposes other than
> building the normal routing table. UPA uses LSInfinity metric only to identify unreachable prefixes, which is contrary to RFC 5305.

no, it is in perfect compliance with RFC 5305.

> [draft-ietf-lsr-ip-flexalgo] also uses LSInfinity metric. It may also be applied to other purposes.

which is orthogonal to our discussion here, because if the valid metric 
exists in IP Algorithm Prefix Reachability Sub-TLV, the metric in parent 
TLV is ignored. There is no problem here.

> Therefore, using LSInfinity metric alone is not enough.

that's what you claim, bit that is not necessary true.

Peter


> In conclusion, the PUA solution is more complete and the unreachable prefix is defined in a more reasonable manner.
> 
> Zhibo Hu
> 
> 
>> Peter
>>
>>
>>>
>>> We have yet mentioned that PUA mechanism has been discussed two years
>>> before the UPA solution.
>>>
>>> More responses are inline. Anyway, I am glad that your comments have
>>> some bases, although you misunderstood something.
>>>
>>> Aijun Wang
>>> China Telecom
>>>
>>>> On Jul 29, 2022, at 02:04, Acee Lindem (acee) <[e-mail]> wrote:
>>>>
>>>> Speaking as WG Member:
>>>>
>>>> Hi Ketan,
>>>>
>>>> Thanks for pointing out the similarities. Even after the recent
>>>> changes,  there are still some difference between the drafts which
>>>> I’ll describe in the baseless comments which follow. For conciseness,
>>>> I’ll refer to the drafts as PUA (Draft Wang) and UPA (Draft Psenak).
>>>>
>>>>   1. Backward Compatibility – Now that PUA has appropriated the metric
>>>>      mechanisms from UPA, it is also backward compatible. However, PUA
>>>>      still proposes extensions the IGPs to advertise the PUA
>>>>      capabilities and says the nodes may misbehave if they don’t agree
>>>>      on these capabilities. I guess removing these was omitted when the
>>>>      UPA metric mechanisms were appropriated.
>>>
>>> WAJ] No. the context in the document just describes why and when the
>>> LSInfinity is necessary. The usages of LSInfinity in two drafts are
>>> different: one(PUA) is to avoid the misbehavior(which is conform to
>>> the RFC rules), another(UPA) is to indicate the unreachable
>>> information(which is not described in the RFC rules)
>>>
>>>> 1.
>>>>   2. Receive Router Behavior – For UPA, the unreachable prefix
>>>>      notification is solely for an event signal to be used by other
>>>>      routers in the IGP domain to trigger actions, e.g., BGP PIC
>>>>      excluding the unreachable prefix.  PUA is used for the switchover
>>>>      of services as well but is also used to modify persistent state.
>>>>      In section 4, the PUA advertisement will trigger the advertisement
>>>>      of the prefix by an ABR that does have a route to the unreachable
>>>>      prefix advertised by another ABR.
>>> [WAJ] Is this one evidence that PUA covers UPA?
>>>>
>>>> 2.
>>>>   3. Advertisement Persistence – PUA is advertised like any other LSA
>>>>      and presumably advertised as long as the prefix is unreachable.
>>>>      Conversely, UPA is an ephemeral LSA that will be withdrawn after
>>>>      enough time is allowed for the event notification to propagate.
>>> [WAJ] No. if there is no network update again, the PUA will not be
>>> advertised “as long as the prefix is unreachable “. Actually, there is
>>> description in the document:
>>>
>>>      “The advertisement of PUAM message should only last one
>> configurable
>>>      period to allow the services that run on the failure prefixes are
>>>      switchovered.  If one prefix is missed before the PUAM takes effect,
>>>      the ABR will not declare its absence via the PUAM.”
>>>
>>>
>>> I think you may ignore them.
>>>
>>>> 3.
>>>>
>>>> In my opinion, UPA is superior to PUA since it is addresses the
>>>> original requirement with minimal overhead and changes to the IGP.
>>>> Even after many revisions, PUA still contains a lot of additional
>>>> unnecessary overhead and complexity. I think the WG should adopt UPA
>>>> and not spend any more time on this discussion.
>>>>
>>> [WAJ] From the above responses, I think you should realize that UPA
>>> just cover very minor part of the overall PUA solution, then your
>>> conclusion should be reverted.
>>> Or else, we can compare these two drafts sentences by sentences.
>>>
>>>> Thanks,
>>>>
>>>> Acee
>>>>
>>>> *From: *Lsr <[e-mail]> on behalf of Ketan Talaulikar
>>>> <[e-mail]>
>>>> *Date: *Thursday, July 28, 2022 at 2:29 AM
>>>> *To: *Aijun Wang <[e-mail]>
>>>> *Cc: *"Les Ginsberg (ginsberg)"
>>>> <ginsberg=[e-mail]>,
>>>> "[e-mail]"
>>>> <[e-mail]>, lsr
>>>> <[e-mail]>
>>>> *Subject: *Re: [Lsr] Comments on
>>>> draft-wang-lsr-prefix-unreachable-annoucement
>>>>
>>>> Hi Aijun,
>>>>
>>>> Indeed, your draft has done a "pivot" in the latest version with the
>>>> use of LSInfinity like the UPA proposal. I hope you will do yet
>>>> another "pivot" to move away from the use of Prefix Originator.
>>>>
>>>> IMHO that would also bring the PUA and UPA proposals much closer to
>>>> each other.
>>>>
>>>> Thanks,
>>>>
>>>> Ketan
>>>>
>>>> On Thu, Jul 28, 2022 at 6:52 AM Aijun Wang <[e-mail]
>>>> <mailto:[e-mail]>> wrote:
>>>>
>>>>      Hi, Les:
>>>>
>>>>      I admire you and your comments as usual, but the baseless comments
>>>>      will decrease your credits within the WG. Would you like to review
>>>>      the update of the draft more carefully, then post your comments?
>>>>      Doing this can avoid misleading some of your followers.
>>>>
>>>>      To facilitate your review, I copied the related contents
>>>>
>> again:(https://datatracker.ietf.org/doc/html/draft-wang-lsr-prefix-unreachable
>> -annoucement-10#section-5
>>>>
>>>> <https://datatracker.ietf.org/doc/html/draft-wang-lsr-prefix-unreacha
>>>> ble-annoucement-10#section-5>)
>>>>
>>>>        If not all of nodes within one area support the PUAM
>>>> capabilities,
>>>>
>>>>         the PUAM message should be advertised with the associated
>>>>      prefix cost
>>>>
>>>>         set to LSInfinity.  According to the description in [RFC2328
>>>>      <https://datatracker.ietf.org/doc/html/rfc2328>],
>>>>
>>>>         [RFC5305 <https://datatracker.ietf.org/doc/html/rfc5305>] and
>>>>      [RFC5308 <https://datatracker.ietf.org/doc/html/rfc5308>], the
>>>>      prefix advertised with such metric value
>>>>
>>>>         will not be considered during the normal SPF computation, then
>>>> such
>>>>
>>>>         additional information will avoid the misbehavior of the nodes
>>>> when
>>>>
>>>>         they don't recognize the PUAM message.
>>>>
>>>>         If all of nodes within one area support the PUAM capabilites,
>>>> the
>>>>
>>>>         PUAM message can be safely advertised without the additional
>>>>
>>>>         LSInfinity metric information.
>>>>
>>>>      Then, how can the “legacy nodes MUST interpret as meaning
>>>>      reachable.” ? I wish to hear your explanation.
>>>>
>>>>      Aijun Wang
>>>>
>>>>      China Telecom
>>>>
>>>>
>>>>
>>>>          On Jul 28, 2022, at 06:39, Les Ginsberg (ginsberg)
>>>>          <ginsberg=[e-mail]
>>>>          <mailto:[e-mail]>> wrote:
>>>>
>>>>          (Preamble: All of what I am going to say I have said many
>>>>          times before – on the list – off the list – in private
>>>>          conversations – in WG meetings…
>>>>
>>>>          I don’t say this to start a discussion with the WG authors –
>>>>          it seems clear that we don’t agree and have no path to
>> agreement.
>>>>
>>>>          My purpose in saying this is to respond to the ongoing
>>>>          existence of this draft and offering my opinion as to what
>>>>          action the WG should take.)
>>>>
>>>>          The mechanism defined in this draft is broken. Not only is it
>>>>          not backwards compatible – the PUA advertisements will be
>>>>          misinterpreted to mean the exact opposite of what is intended
>>>>          i.e., the intent is to signal that a prefix is unreachable,
>>>>          but you do so by using an advertisement which legacy nodes
>>>>          MUST interpret as meaning reachable. This is simply broken and
>>>>          should not be done.
>>>>
>>>>          The authors deserve credit for bringing the attention of the
>>>>          WG to the problem space – but the solution offered is not
>>>>          deployable. Given the long period of time during which this
>>>>          draft has been published and the many times it has been
>>>>          presented/discussed in the WG I think it is now time to say
>>>>          thank you to the authors for their work, but the WG is not
>>>>          interested in adopting this draft.
>>>>
>>>>             Les
>>>>
>>>>          *From:* Lsr <[e-mail]
>>>>          <mailto:[e-mail]>> *On Behalf Of *Ketan Talaulikar
>>>>          *Sent:* Wednesday, July 27, 2022 1:36 AM
>>>>          *To:* [e-mail]
>>>>
>> <mailto:[e-mail]>
>>>>          *Cc:* lsr <[e-mail] <mailto:[e-mail]>>
>>>>          *Subject:* [Lsr] Comments on
>>>>          draft-wang-lsr-prefix-unreachable-annoucement
>>>>
>>>>          Hello Authors,
>>>>
>>>>          I am sharing some comments on the latest version of this
>>>>          document since we seem to have a packed agenda in LSR this
>> time.
>>>>
>>>>          1) I notice that in the latest update of the draft, there is a
>>>>          big change to start using LSInfinity for indicating prefix
>>>>          unreachability (similar
>>>>          to draft-ppsenak-lsr-igp-ureach-prefix-announce). I see this
>>>>          as a sign of a degree of convergence between the two drafts.
>>>>
>>>>          2) However, I then question the motivation of the authors to
>>>>          continue with the bad design of overloading Prefix Originator
>>>>          and the added capability stuff on top. I don't wish to repeat
>>>>          why that design was an absolute NO-GO for me and I am glad to
>>>>          see the authors acknowledge the problem with misrouting by
>>>>          implementations not supporting this specification. So I don't
>>>>          see the point of still retaining all that.
>>>>
>>>>          Thanks,
>>>>
>>>>          Ketan
>>>>
>>>>          _______________________________________________
>>>>          Lsr mailing list
>>>>          [e-mail] <mailto:[e-mail]>
>>>>          https://www.ietf.org/mailman/listinfo/lsr
>>>>          <https://www.ietf.org/mailman/listinfo/lsr>
>>>>
>>>> _______________________________________________
>>>> Lsr mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/lsr
>>
>> _______________________________________________
>> Lsr mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/lsr
>