Re: [Lsr] Rtgdir last call review of draft-ietf-isis-te-app-13

["Les Ginsberg (ginsberg)" <[e-mail]>]

Bruno -

Thanx again for your review.
V14 of the draft has been posted to address your comments.

Please let me know if you believe there are still outstanding issues.

A few more remarks inline.

> -----Original Message-----
> From: [e-mail] <[e-mail]>
> Sent: Tuesday, June 02, 2020 9:33 AM
> To: Les Ginsberg (ginsberg) <[e-mail]>
> Cc: [e-mail]; [e-mail]; [e-mail]; rtg-
> [e-mail]
> Subject: RE: Rtgdir last call review of draft-ietf-isis-te-app-13
> 
> Les,
> 
> Thanks for your answers.
> Comments inline
> 
> > From: Les Ginsberg (ginsberg) [mailto:[e-mail]]
> > Sent: Saturday, May 30, 2020 12:09 AM
> >
> > Bruno -
> >
> > Thanx for your (as always) meticulous review.
> > Responses inline.
> > Once we have reached agreement I will send out an updated version.
> >
> > > -----Original Message-----
> > > From: Bruno Decraene via Datatracker <[e-mail]>
> > > Sent: Friday, May 29, 2020 10:18 AM
> > > To: [e-mail]
> > > Cc: [e-mail]; [e-mail]; [e-mail]
> > > Subject: Rtgdir last call review of draft-ietf-isis-te-app-13
> > >
> > > Reviewer: Bruno Decraene
> > > Review result: Has Issues
> > >
> > >  Hello,
> > >
> > > I have been selected as the Routing Directorate reviewer for this draft.
> The
> > > Routing Directorate seeks to review all routing or routing-related drafts as
> > > they pass through IETF last call and IESG review, and sometimes on
> special
> > > request. The purpose of the review is to provide assistance to the
> Routing
> > > ADs.
> > > For more information about the Routing Directorate, please see
> > > http://trac.tools.ietf.org/area/rtg/trac/wiki/RtgDir
> > >
> > > Although these comments are primarily for the use of the Routing ADs, it
> > > would
> > > be helpful if you could consider them along with any other IETF Last Call
> > > comments that you receive, and strive to resolve them through
> discussion or
> > > by
> > > updating the draft.
> > >
> > > Document: draft-ietf-isis-te-app-13
> > > Reviewer: Bruno Decraene
> > > Review Date: 2020-05-29
> > > IETF LC End Date: 2020-05-29
> > > Intended Status: Standards Track
> > >
> > > Summary:
> > >     I have some minor concerns about this document that I think should be
> > >     resolved before publication.
> > >
> > > Comments:
> > >   Draft is clear.
> > >
> > > Minor Issues:
> > >
> > > §4.1
> > > *2 (for SABM & UDABM fields)
> > > OLD: The length SHOULD be the minimum required to send all bits which
> are
> > > set.
> > > I'd propose
> > > NEW: The length SHOULD be the minimum required to send all the
> > > meaningful bits
> > > which are set.
> > >
> > > Motivation; the 'bits which are sent' are the bits in the SABM field. (they
> do
> > > include non-meaningful and padding bits)
> > >
> >
> > [Les:] The definition of what is "meaningful" and what is "padding"  to me is
> ambiguous.
> > Meaningful could be only those bits which are currently defined in the
> registry (speaking of SABM here). But if there are 10 bits defined in the
> registry and I only intend to set Bit 5, I do not need to send all 10 bits - I only
> need to send one octet - because we state:
> >
> > "Bits that are NOT transmitted MUST be treated as if they
>    > are set to 0 on receipt.  "
> >
> > Also, an implementation written when there were only 4 bits defined in
> the registry might think that "meaningful" is different than an
> implementation written when more than 8 bits were defined in the registry.
> Yet they can still interoperate.
> >
> > I believe the current language is best.
> 
> [Bruno]
> I withdraw my comment. Sorry for the noise.
> I had read "bits which are sent", while the text is "bits which are set".
> 
> 
> > > ----
> > >
> > > OLD: Undefined bits MUST be transmitted as 0
> > > NEW: Undefined transmitted bits MUST be cleared (0)
> > >
> > > Motivation: currently the number of undefined bits is 8*8-3. They
> SHOULD
> > > not be
> > > transmitted (beyond the first ones fitting in the first N required octet).
> The
> > > sentence "Undefined bits MUST be transmitted as 0" could be read as all
> > > defined
> > > bits MUST be transmitted (as 0).
> > > ---
> > [Les:] I do not see how that could be a valid interpretation given that we
> state:
> >
> > " The length SHOULD  be the minimum required to send all bits which are
> set."
> 
> [Bruno]
> So we have
> 1) The length SHOULD  be the minimum required to send all bits which are
> set
> 2) Undefined bits MUST be transmitted as 0
> 
> Given the "MUST"  vs "SHOULD" and "transmitted" (which means "sent"), I
> do believe my proposal is better. But I won't insist.
> 

[Les:] I took a second look at this and appreciated your point better.
I changed the text to read:

" Undefined bits which are transmitted MUST be transmitted as 0..."

> 
> > And (repeating)
> >
> > "Bits that are NOT transmitted MUST be treated as if they
>    > are set to 0 on receipt.  "
> >
> > And again, you assume that "defined bits" is the same for all
> implementations - which isn't guaranteed as I discussed above.
> 
> [Bruno] I don't think that this matter as the behavior is specific to the sender.
> In addition, the term " Undefined bits" is yours.
> 
> >
> > > User Defined Application Identifier Bits have no name. I'd propose to call
> > > them
> > > UDABM[0], UDABM[1]... This may avoid that different implementation
> use
> > > different names and, more problematic, that some implementations
> starting
> > > with
> > > 1 (the first, the second) while while some other implementations starts
> as 0,
> > > creating interop issues (SABM[1] on node A is SABM[0] on node B)
> > > ---
> >
> > [Les:] What implementations may name bits they assign from the User
> space is out of scope of this document.
> > If I were implementing a non-standard User App I likely would give it a
> meaningful name both in my code and in any documentation I produce.
> 
> [Bruno] ok, let's leave the terminology choice for this parameter to an
> hypothetical yang model.
> 
> > As far as interoperability, if you want multiple vendors to interoperate then
> you need a standard application. User defined applications do not provide
> any guarantee of interoperability.
> >
> > We do state that
> >
> > "It is recommended that [user defined] bits are used starting with Bit 0..."
> >
> > but as User Defined Applications are outside the scope of the document
> they might choose to do otherwise.
> >
> >
> > > §4.2
> > >
> > > "In cases where conflicting values for the same application/attribute/link
> are
> > > advertised all the conflicting values MUST be ignored." I'd propose to add
> > > "for
> > > this application" (IOW, those values are still applicable for all other
> > > applications)
> > > ---
> >
> > [Les:] How about adding "for the specified application" ?
> 
> [Bruno] Looks good.
> 
> 
> > > §6.2
> > > I'd argue that the first part of section 3.2 is a specification of the behavior
> > > and hence should be moved to section 4.1, rather than placed in the
> section
> > > "deployment consideration" which eventually will not be read by
> someone
> > > implementing the specification. Especially since the text in section 4.1
> > > implies a different behavior: "Bits that are NOT transmitted MUST be
> treated
> > > as
> > > if they are set to 0 on receipt."
> > > ---
> >
> > [Les:] I think you meant to say the "first part of section 6.2"?? Correct?
> 
> [Bruno] yes, you are correct.
> 
> >
> > If so, I agree - and will move that text - though I would prefer to put it into
> Section 4.2.
> > Section 4.1 is describing the encoding of the bit mask. Section 4.2 describes
> the ASLA sub-TLV and how to interpret it.
> > For example, that is where L-bit is discussed.
> > Sound good to you?
> 
> [Bruno] Looks good. Thank you.
> 
> >
> > > §5
> > > "In the case of SRTE, advertisement of application specific link attributes
> > > does NOT indicate enablement of SRTE." What does "enablement of
> SRTE"
> > > means? Do
> > > you have a pointer to a document/text?
> > >
> > > I'm not sure I would keep that paragraph on SR-TE enablement.
> > > ---
> >
> > [Les:] The paragraph is required because we state
> >
> > "the relationship between application specific link attribute
>    > advertisements and enablement for that application"
> >
> > is required for all new applications.
> 
> [Bruno]   The argument seems weak to me. Change "MUST" to "SHOULD"
> and voilà, problem solved!
> Also the requirement is 'for the future' and to be defined application. Stricto
> census it does not apply to you in this draft.
> 
> > In this document we are providing that definition for the three existing
> applications.
> >
> > The paragraph does state:
> >
> > "SRTE is implicitly enabled on all links
>    > which are part of the Segment Routing enabled topology independent of
>    > the existence of link attribute advertisements."
> >
> > I will modify the first sentence to say:
> >
> > "In the case of SRTE, advertisement of application specific link
>    > attributes does NOT indicate enablement of SRTE  on that link."
> >
> > ("on that link" is added)
> >
> > Does this work for you?
> 
> [Bruno]   I still have the same question: What does "enablement of SRTE"
> means?
> 
[Les:] As stated in the draft, 

" SRTE is implicitly enabled
   on all links which are part of the Segment Routing enabled topology
   independent of the existence of link attribute advertisements."

This means that all links in an SR enabled topology may be used by SRTE. Link attribute advertisements serve to provide information which can be used to apply constraints, but they are not necessary in order for the link to be used as part of an SR Policy.
The most obvious example of this is a policy composed of adjacency-SIDs, directing the traffic along a specific set of links independent of any advertised link attributes.
HTH.

> 
> > > §6.1
> > > "Under the conditions defined above, implementations which support
> the
> > >    extensions defined in this document have the choice of using legacy
> > >    advertisements or application specific advertisements in support of
> > >    SRTE and/or LFA.  This will require implementations to provide
> > >    controls specifying which type of advertisements are to be sent/
> > >    processed on receive for these applications."
> > >
> > > I think that "have the choice" is not prescriptive enough given the
> > > deployment
> > > issues described in section 6.3 I'd rather say that implementations MUST
> > > support the use of both advertisements (legacy and application specific
> > > advertisement) and MUST provide controls specifying which type of
> > > advertisements are to be processed on receive for these applications.
> > >
> >
> > [Les:] We know that existing deployments (pre-this draft) use legacy for
> SRTE/LFA.
> > In the future, implementations could choose to migrate to using the new
> ASLA advertisements for SRTE/LFA. Whether they will do so or not is a
> business decision.
> 
> [Bruno] As written in the draft, this is required for interop. So I don't see this
> as a business decision
> 
> I'm quoting the draft
> in section 6.3 "deployments using the
>    extensions defined in this document must be able to co-exist with use
>    of the legacy advertisements by routers which do not support the
>    extensions defined in this document."
> 
> In order for deployments to be able to follow this 'must', the implementation
> MUST support it.
> 
> In section 6.3.1 "interoperability is achieved by using legacy advertisements
> and
>    sending application specific advertisements with L-flag set and no
>    link attribute values."
> 
> In section 6.3.3
> "So long as there is any
>    legacy router in the network which has any of the applications
>    enabled, all routers MUST continue to advertise link attributes using
>    legacy advertisements."
> 
> So from above, all routers MUST be capable of sending and receiving legacy
> advertisements. This seem to be aligned with my text.
> 
[Les:] Multiple deployment scenarios are possible.
There may be a deployment where legacy routers and routers supporting the extensions defined in this draft are present and SRTE (for example) is in use. In this case it is necessary that the updated routers be able to support legacy advertisements.

But there may also be a deployment where only upgraded routers are deployed and SRTE is in use. In this case support of legacy advertisements is NOT required.

Vendors may make the decision - now or in the future - to deprecate support for legacy advertisements in their implementations. Clearly, if they do so they will not be able to interoperate with legacy routers. But if they do not see such a limitation as "bad for business" then they may opt to do that.
This does not make these implementations in violation of this specification - which is why using language which requires implementations to always support legacy is inappropriate.

I did modify a sentence in Section 6.3 to say

" Therefore deployments using the
   extensions defined in this document in the presence of routers which
   do not support these extensions need to be able to interoperate with
   the use of legacy advertisements by the legacy routers."

   Les

> --Bruno
> 
> > We do not want to declare implementations as non-conformant if they do
> not migrate.
> 
> 
> 
> 
> >
>    > Les
> >
> 
> __________________________________________________________
> __________________________________________________________
> _____
> 
> Ce message et ses pieces jointes peuvent contenir des informations
> confidentielles ou privilegiees et ne doivent donc
> pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce
> message par erreur, veuillez le signaler
> a l'expediteur et le detruire ainsi que les pieces jointes. Les messages
> electroniques etant susceptibles d'alteration,
> Orange decline toute responsabilite si ce message a ete altere, deforme ou
> falsifie. Merci.
> 
> This message and its attachments may contain confidential or privileged
> information that may be protected by law;
> they should not be distributed, used or copied without authorisation.
> If you have received this email in error, please notify the sender and delete
> this message and its attachments.
> As emails may be altered, Orange is not liable for messages that have been
> modified, changed or falsified.
> Thank you.