Re: [spring] Spring protection - determining applicability

[Martin Horneffer <[e-mail]>]

A few thoughts from my (operator's) PoV:

  - The disussion is a very good and important one. It probably should 
be discussed and documented well in order to justify the proposed 
protections mechanisms.

  - Not all operators seem to have the same requirements.
     (A somewhat similar discussion might be the one for disjoint paths. 
Those are often equired by voice signalling applications. In some cases 
the voice service demands that traffic is blackholed rather than on 
forwarded on the wrong path. In other cases disjoint paths are just 
required for the "good case". Traffic MAY be forwarded on the wrong 
path, as long as the network just makes sure the traffic on the other 
path is never affected by the same failure.)

  - Personally I would hate to see yet another IGP extension for this 
purpose.

  - I would rather prefer a good discussion of what can be achieved by 
using easy to make switches:
     - The protection behaviour could be switched on or off per node.
        - An operator with strict "some traffic may never touch certain 
parts of the network" requirements might switch off the behaviour, while 
others might switch it on.
     - A node could allow a switch even individually for every port or 
neighbor.
        - If a node knows that one of it's neighbours is a service node 
rather than a plain topological one, it could switch off protection. 
This is how I would prefer to solve the problem with serrvice nodes.

Should this be discussed in the protection document, or in a separate one?

Best regards, Martin


Am 14.08.20 um 19:45 schrieb Ketan Talaulikar (ketant):
>
> Hi Robert,
>
> We do not have a signalling mechanism in IGPs today to indicate a 
> “bypass-able” indication for Prefix SIDs. If there was a desire for 
> it, an IGP extension would be required (there is none in progress 
> AFAIK). Note that this results in doubling the prefix SID scale 
> (global labels) in the network. So I would not go about this trivially.
>
> I think it helps to get more inputs and perspectives from operators on 
> their views for doing a bypass via local protection for segments in an 
> SR Policy. There may be those that prefer end-to-end path protection 
> using a fallback path that is say disjoint with the primary but 
> provides an appropriate SLA/intent?
>
> Thanks,
>
> Ketan
>
> *From:*Robert Raszuk <[e-mail]>
> *Sent:* 14 August 2020 23:04
> *To:* Ketan Talaulikar (ketant) <[e-mail]>
> *Cc:* Alexander Vainshtein <[e-mail]>; Joel M. 
> Halpern <[e-mail]>; Shraddha Hegde <[e-mail]>; 
> [e-mail] <[e-mail]>; 
> [e-mail]
> *Subject:* Re: [spring] Spring protection - determining applicability
>
> Ketan,
>
> Looks like we are pretty much in sync here.
>
> But let me just observe that I purposely did not mention about SR 
> policies as we are not able to signal the intent with the packets itself.
>
> So all we have there is SIDs. BSIDs or prefix SIDs need to be flooded 
> with information if policies build with using them are bypass eligible 
> or not.
>
> I was actually under the impression that this is already there and I 
> am just not aware, but looking deeper indeed I do not see this marking 
> neither in ISIS nor OSPF for prefix SIDs.
>
> Is there some work in progress to add it to those protocols or have we 
> just documented need for a short LSR draft  ?
>
> Thx,
>
> R.
>
> On Fri, Aug 14, 2020 at 6:17 PM Ketan Talaulikar (ketant) 
> <[e-mail] <mailto:[e-mail]>> wrote:
>
>     Hi Robert,
>
>     Please check inline below.
>
>     *From:*Robert Raszuk <[e-mail] <mailto:[e-mail]>>
>     *Sent:* 14 August 2020 21:13
>     *To:* Ketan Talaulikar (ketant) <[e-mail]
>     <mailto:[e-mail]>>
>     *Cc:* Alexander Vainshtein <[e-mail]
>     <mailto:[e-mail]>>; Joel M. Halpern
>     <[e-mail] <mailto:[e-mail]>>; Shraddha Hegde
>     <[e-mail] <mailto:[e-mail]>>;
>     [e-mail]
>     <mailto:[e-mail]>
>     <[e-mail]
>     <mailto:[e-mail]>>; [e-mail]
>     <mailto:[e-mail]>
>     *Subject:* Re: [spring] Spring protection - determining applicability
>
>     Hi Ketan,
>
>     While I completely agree with your note the consequences of it are
>     pretty sevre.
>
>     */[KT] I understand. We need to be mindful of implications of
>     protection schemes for the SLAs/intent of SR Policies./*
>
>     Unless we signal which prefix SID is protection eligible and which
>     is not how would other nodes know if they can protect it or not ?
>
>     */[KT] Correct. To be more accurate, we need to consider this more
>     in the context of SLA or “intent” of SR Policies and which
>     segments may be “bypass-able” for local protection for some of
>     those SR Policies. We also have path-protection mechanisms./*
>
>     It seems that today's safe thing is not to apply any node
>     protection on SR flows at the PLRs then.
>
>     And link protection MUST assure that packets will arrive at the
>     neighbor node via some other link regardless of further path
>     towards destination.
>
>     */[KT] Yes. We have a mechanism to indicate which adj-SIDs have
>     protection (that mechanism only provides link protection to get to
>     the neighbor node) so the SR Policy computation is able to
>     indicate whether that specific link is “bypass-able” or not by its
>     choice of protected or unprotected adj-SIDs respectively./*
>
>     *//*
>
>     */Thanks,/*
>
>     */Ketan/*
>
>     Is it correct ?
>
>     Thx
>
>     R
>
>     On Fri, Aug 14, 2020 at 5:32 PM Ketan Talaulikar (ketant)
>     <[e-mail] <mailto:[e-mail]>> wrote:
>
>         Hi Sasha,
>
>         The service node advertises its own Prefix SID. The service
>         function that this service node implements does not require
>         any context (i.e. all packets arriving at the node are
>         subjected to that service). Therefore the service node does
>         not need to receive a packet with it’s own Prefix SID.
>
>         Thus, we cannot assume that when PHP is used, then the SID is
>         only associated with a topological instruction.
>
>         Hope that clarifies?
>
>         Thanks,
>
>         Ketan
>
>         *From:*Alexander Vainshtein <[e-mail]
>         <mailto:[e-mail]>>
>         *Sent:* 14 August 2020 20:24
>         *To:* Ketan Talaulikar (ketant) <[e-mail]
>         <mailto:[e-mail]>>; Joel M. Halpern
>         <[e-mail] <mailto:[e-mail]>>; Shraddha
>         Hegde <[e-mail] <mailto:[e-mail]>>;
>         [e-mail]
>         <mailto:[e-mail]>
>         <[e-mail]
>         <mailto:[e-mail]>>; Robert Raszuk
>         <[e-mail] <mailto:[e-mail]>>
>         *Cc:* [e-mail] <mailto:[e-mail]>
>         *Subject:* Re: [spring] Spring protection - determining
>         applicability
>
>         Ketan, and all,
>
>         I have stated that, IMHO and FWIW, both Adj-SIDs and Prefix
>         SIDs that are advertised with PHP can  ONLY represent
>         topological instructions in SR-MPLS - because the advertising
>         node will not receive them and therefore can hardly be
>         expected to associate any service function with them.
>
>         This is complementary to what you have said.
>
>         Hope this clarifies my position.
>
>         What, if anything, did I miss?
>
>         Regards,
>
>         Sasha
>
>         Get Outlook for Android <https://aka.ms/ghei36>
>
>         ------------------------------------------------------------------------
>
>         *From:* Ketan Talaulikar (ketant) <[e-mail]
>         <mailto:[e-mail]>>
>         *Sent:* Friday, August 14, 2020, 16:23
>         *To:* Alexander Vainshtein; Joel M. Halpern; Shraddha Hegde;
>         [e-mail]
>         <mailto:[e-mail]>; Robert Raszuk
>         *Cc:* [e-mail] <mailto:[e-mail]>
>         *Subject:* RE: [spring] Spring protection - determining
>         applicability
>
>         ------------------------------------------------------------------------
>
>         NOTICE: This email was received from an EXTERNAL sender
>
>         ------------------------------------------------------------------------
>
>         Hi Sasha,
>
>         If the service does not need any additional context (e.g. a
>         firewall that just applies locally configured default rules on
>         it), then I don’t see why PHP could not be done for a Prefix
>         SID associated with a service node.
>
>         Also, I didn’t follow the point that you were trying to make
>         about Adj-SIDs.
>
>         Thanks,
>
>         Ketan
>
>         *From:*Alexander Vainshtein <[e-mail]
>         <mailto:[e-mail]>>
>         *Sent:* 14 August 2020 18:24
>         *To:* Ketan Talaulikar (ketant) <[e-mail]
>         <mailto:[e-mail]>>; Joel M. Halpern
>         <[e-mail] <mailto:[e-mail]>>; Alexander
>         Vainshtein <[e-mail]
>         <mailto:[e-mail]>>; Shraddha Hegde
>         <[e-mail] <mailto:[e-mail]>>;
>         [e-mail]
>         <mailto:[e-mail]>
>         <[e-mail]
>         <mailto:[e-mail]>>; Robert Raszuk
>         <[e-mail] <mailto:[e-mail]>>
>         *Cc:* [e-mail] <mailto:[e-mail]>
>         *Subject:* Re: [spring] Spring protection - determining
>         applicability
>
>         Hi all,
>
>         Regarding the statement "Prefix SID could be just a
>         topological instruction or may also be used to steer the flow
>         to a node which is applying a service function to it":
>
>         I think that in SR-MPLS a Node SID that is advertised with PHP
>         aciton can be safely considered as "just a topological
>         instruction" by the PLR because the originating node will not
>         receive it.
>
>         The same applies to Adj-SDIs.
>
>         My 2c.
>
>         Get Outlook for Android
>         <https://clicktime.symantec.com/375c5YYBeEbaEZwUcHpCY1m6H2?u=https%3A%2F%2Faka.ms%2Fghei36>
>
>         ------------------------------------------------------------------------
>
>         *From:* spring <[e-mail]
>         <mailto:[e-mail]>> on behalf of Ketan
>         Talaulikar (ketant) <ketant=[e-mail]
>         <mailto:ketant=[e-mail]>>
>         *Sent:* Friday, August 14, 2020, 15:00
>         *To:* Joel M. Halpern; Alexander Vainshtein; Shraddha Hegde;
>         [e-mail]
>         <mailto:[e-mail]>; Robert Raszuk
>         *Cc:* [e-mail] <mailto:[e-mail]>
>         *Subject:* Re: [spring] Spring protection - determining
>         applicability
>
>         Hi All,
>
>         I would like to share a different perspective on this.
>
>         First, thanks to Joel for bringing up the discussion. Clearly
>         we need a well-defined applicability statement for determining
>         applicability of protection for segment used in an SR Policy.
>         Some of this is captured in [1].
>
>         This is about local repair at a PLR. By it's very nature, the
>         PLR does not have a notion of how "strict or not" is the SLA
>         that is being provided by the SR Policy. Awareness of that
>         notion exists at the SR Policy headend and/or computation-node.
>
>         We have protected and un-protected variants of adjacency SIDs
>         to enable the computation to pick or the other based on the
>         "strictness" of the SLA requirement for picking that link. We
>         do not have such a notion for Prefix SIDs. One can say that we
>         could introduce signalling (e.g. a B flag) to indicate whether
>         a Prefix SID can be bypassed or not. This provides the
>         opportunity for the computation to use one or the other flavor
>         depending on the nature of the SLA for the SR Policy.
>
>         I have a problem and a concern in the assumption that PLRs can
>         assume that the currently defined variant of Prefix SIDs in
>         RFC8402 (and IGP specs) are "bypass-able".
>
>         As Joel and others have brought out, the Prefix SID could be
>         just a topological instruction or may also be used to steer
>         the flow to a node which is applying a service function to it.
>         In order to support a mix of SR Policies of different SLAs
>         (strict and not-strict), we need to enable the choice of SIDs
>         that indicates to the PLR whether they are "bypass-able" or not.
>
>         For the cases, where the SR Policy has a specific SLA, it is
>         required for nodes to drop the packets meant for the "active
>         segment" than to bypass it. When this mechanism is used along
>         side SRTE path monitoring mechanisms, it enables the headend
>         to detect the failure and fallback to an alternate path using
>         the path protection approach. This is something that is
>         described and in use in deployments today [1]..
>
>         Thanks,
>         Ketan
>
>         [1]
>         https://clicktime.symantec.com/3Y3fWuNFYCjMJUiiAiWwUms6H2?u=https%3A%2F%2Ftools.ietf.org%2Fhtml%2Fdraft-ietf-spring-segment-routing-policy-08%23section-9
>         [2]
>         https://clicktime.symantec.com/36fCMrgmEewC4a4AJRrmn7H6H2?u=https%3A%2F%2Ftools.ietf.org%2Fhtml%2Fdraft-ietf-spring-segment-routing-policy-08%23section-9.3
>
>         -----Original Message-----
>         From: spring <[e-mail]
>         <mailto:[e-mail]>> On Behalf Of Joel M. Halpern
>         Sent: 04 August 2020 20:25
>         To: Alexander Vainshtein <[e-mail]
>         <mailto:[e-mail]>>; Shraddha Hegde
>         <shraddha=[e-mail]
>         <mailto:shraddha=[e-mail]>>;
>         [e-mail]
>         <mailto:[e-mail]>
>         <[e-mail]
>         <mailto:[e-mail]>>; Robert Raszuk
>         <[e-mail] <mailto:[e-mail]>>
>         Cc: [e-mail] <mailto:[e-mail]>; Joel M. Halpern
>         <[e-mail] <mailto:[e-mail]>>
>         Subject: Re: [spring] Spring protection - determining
>         applicability
>
>         There are, as far as I can tell, a number of ways to address
>         this family of related questions.
>         What struck me, and prompted the starting question, was that
>         none of them were spelled out.  I see lots of interesting
>         ideas / proposals.
>         Some of them are compatible with others.   Some are not.
>         It would be good if we could reach agreement on how we thought
>         it should be handled.
>
>         Thank you,
>         Joel
>
>         On 8/4/2020 3:54 AM, Alexander Vainshtein wrote:
>         > Hi all,
>         >
>         > I am still not sure that the problem of bypass going thru
>         undesirable
>         > links/nodes exists in the case of topological SIDs.
>         >
>         > AFAIK, Facility Protection in RSVP-TE FRR (RFC 4090
>         >
>         <https://clicktime.symantec.com/3Q92knE9XJujrf8Bk7oJvs6H2?u=https%3A%2F%2Ftools.ietf.org%2Fhtml%2Frfc4090>)
>         has been successfully deployed
>         > for many years before SR-MPLS has been introduced. What’s more,
>         > signaling of bypass tunnels he PLR usually did not include
>         any of the
>         > constraints used for computing of any specific LSP that the
>         bypass LSP
>         > would protect – because in the Facility Protection mode the
>         same
>         > bypass LSP would be used to protect multiple LSPs passing
>         thru the
>         > failed link/node.
>         >
>         >  From my POV the only difference between this behavior and that
>         > introduced by the “bypassing” drafts in SR is that, in the
>         case of
>         > RSVP-TE, the operator would explicitly indicate, as part of LSP
>         > signaling, whether it would or would not use FRR; LSPs that
>         would not
>         > use FRR would then drop traffic rather than delivering it
>         the wrong way.
>         >
>         > Such an option indeed does not exist in SR-TE today, but
>         would be easy
>         > to provide if so desired IMHO.
>         >
>         > Did I miss something substantial?
>         >
>         > Regards, and lots of thanks in advance,
>         >
>         > Sasha
>         >
>         > Office: [phone]
>         >
>         > Cell:      [phone]
>         >
>         > Email: [e-mail]
>         <mailto:[e-mail]>
>         >
>         > *From:* spring <[e-mail]
>         <mailto:[e-mail]>> *On Behalf Of *Shraddha Hegde
>         > *Sent:* Tuesday, August 4, 2020 9:41 AM
>         > *To:* [e-mail]
>         <mailto:[e-mail]>
>         > <[e-mail]
>         <mailto:[e-mail]>>; Robert Raszuk
>         <[e-mail] <mailto:[e-mail]>>
>         > *Cc:* [e-mail] <mailto:[e-mail]>; Joel M.
>         Halpern <[e-mail] <mailto:[e-mail]>>
>         > *Subject:* Re: [spring] Spring protection - determining
>         applicability
>         >
>         > All,
>         >
>         > This is a very interesting discussion and thanks to Joel for
>         starting
>         > this discussion. IMO, when there are strict requirements of
>         avoiding
>         > certain nodes/links it can be realized  either by defining a
>         flex-algo
>         > avoiding those
>         >
>         > Nodes and links or by using a stack of unprotected adj-sids
>         that avoid
>         > restricted nodes and links. When a stack of adj-sids is used to
>         > realize the path, the head-end based (sBFD) protection
>         mechanisms can be applied.
>         >
>         > If Node-sids/prefix-sid/anycast-sids are used to build the
>         stack, the
>         > failure events may cause traffic to go through restricted
>         nodes and
>         > links. This would happen regardless of whether any kind of
>         protection
>         > is in use or not.
>         >
>         > Rgds
>         >
>         > Shraddha
>         >
>         > Juniper Business Use Only
>         >
>         > *From:* spring <[e-mail]
>         <mailto:[e-mail]%20%0b>>
>         <mailto:[e-mail]>> *On Behalf Of *Andrew Alston
>         > *Sent:* Tuesday, August 4, 2020 5:41 AM
>         > *To:* Robert Raszuk <[e-mail]
>         <mailto:[e-mail]> <mailto:[e-mail]>>
>         > *Cc:* [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]>; Joel M. Halpern
>         > <[e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]>>
>         > *Subject:* Re: [spring] Spring protection - determining
>         applicability
>         >
>         > *[External Email. Be cautious of content]*
>         >
>         > Robert this is actually far more difficult when – it can be
>         an entire
>         > (long) series of nodes that need to be avoided.
>         >
>         > It could potentially be made to work but I’d worry that to
>         do this –
>         > you’d have to stack 10 – 20 – 30 negative labels – and that
>         wouldn’t
>         > be viable.
>         >
>         > It’s easier to use algorithms and adjacency sids and other
>         such things
>         > to calculate paths – the biggest trick is about the stack
>         depth.  When
>         > you have this need for node avoidance – the need for 10+
>         label depth
>         > is critical – unless you wanna be applying one hell of a lot of
>         > binding labels along the way which is a nightmare.
>         >
>         > But to answer your question, is this a common use case –
>         it’s a use
>         > case that most of the people I discuss this with certain
>         have – I cant
>         > comment on a global scale, or for anyone else, but every
>         indication I
>         > have is that yes – its something people need, and want
>         >
>         > Andrew
>         >
>         > *From:* Robert Raszuk <[e-mail]
>         <mailto:[e-mail]> <mailto:[e-mail]>>
>         > *Sent:* Tuesday, 4 August 2020 01:27
>         > *To:* Andrew Alston <[e-mail]
>         <mailto:[e-mail]%20%0b>>
>         <mailto:[e-mail]>>
>         > *Cc:* Joel M. Halpern <[e-mail]
>         <mailto:[e-mail]%20%0b>>
>         <mailto:[e-mail]>>; [e-mail]
>         <mailto:[e-mail]>
>         > <mailto:[e-mail]>
>         > *Subject:* Re: [spring] Spring protection - determining
>         applicability
>         >
>         > Is this a common use case ie. "but rather – which nodes /
>         network
>         > segments it can never touch or flow through."
>         >
>         > If so perhaps its time to define notion of *negative-SID*
>         ie. list in
>         > the packet resources which given packet MUST not ever traverse.
>         >
>         > Put in the packet set of nodes or links which the packet
>         should never
>         > traverse.
>         >
>         > That goes in line of recent wave of negative routing
>         implementations
>         > (RIFT) or discussions (LSR)
>         >
>         > Best,
>         > R.
>         >
>         > On Mon, Aug 3, 2020 at 11:46 PM Andrew Alston
>         > <[e-mail]
>         <mailto:[e-mail]%20%0b>>
>         <mailto:[e-mail]>> wrote:
>         >
>         >     So –
>         >
>         >     One of the use cases, in fact, some very major use cases
>         in any
>         >     spring technology for us revolve around the following
>         >
>         >     a.The explicit avoidance of certain nodes
>         >
>         >     b.The explicit avoidance of certain sections of the network
>         >
>         >     Anything that could result in that explicit avoidance
>         being violated
>         >     – would create, shall we say significant problems.
>         >
>         >     Much of the use case is not a case of which nodes the
>         packets flow
>         >     through – but rather – which nodes / network segments it
>         can never
>         >     touch or flow through. Effectively, to be used as a
>         technology to
>         >     avoid certain things for specific reasons.
>         >
>         >     This is also one of the reasons for needing such deep
>         label stacks –
>         >     this kind of detailed path programming tends to deepen
>         the stack
>         >     because you sometimes have to be pretty explicit.
>         >
>         >     It is absolutely critical to us that this functionality
>         is there –
>         >     and that we can avoid situations which could cause
>         traffic to
>         >     accidently hit things explicitly avoided.
>         >
>         >     I wish I could be more specific than this, but it is
>         what it is.
>         >
>         >     Thanks
>         >
>         >     Andrew
>         >
>         >     *From:* spring <[e-mail]
>         <mailto:[e-mail]%0b>>    
>         <mailto:[e-mail]>> *On Behalf Of *Joel M. Halpern
>         >     *Sent:* Monday, 3 August 2020 21:36
>         >     *To:* Robert Raszuk <[e-mail]
>         <mailto:[e-mail]> <mailto:[e-mail]>>
>         >     *Cc:* spring@ietf..org <mailto:spring@ietf..org>
>         <mailto:[e-mail]>
>         >     *Subject:* Re: [spring] Spring protection - determining
>         > applicability
>         >
>         >     (Since the thread has gotten long enough, reiterating
>         that this is as a
>         >     participant, not a WG chair.)
>         >
>         >     Yes, we are talking IP networks. And yes, I have seen IP
>         networks that
>         >     choose to drop packets. For all sorts of reasons.
>         >     I think there are likely other reasons why one may not
>         want a random
>         >     path rather than a chosen TE path. I think it is
>         important we be clear
>         >     about what constraints may be / are violated when we
>         tell people they
>         >     have this tool (protective rerouting) that is intended
>         to preserve QoS.
>         >
>         >     Let's be clear. I am not arguing that this is not a good
>         idea. It is a
>         >     good idea. And useful. I am trying to figure otu what
>         combination of
>         >     additional mechanisms and clear descriptions will lead
>         to everyone
>         >     getting the behavior they expect (which may not be the
>         behavior they
>         >     desire, but sometimes is the best we can do.)
>         >
>         >     Yours,
>         >     Joel
>         >
>         >     On 8/3/2020 2:30 PM, Robert Raszuk wrote:
>         >      > Joel,
>         >      >
>         >      > Are we still talking about IP networks here ? Or
>         perhaps some hard
>         >      > slicing with real resource reservations or detnets ?
>         >      >
>         >      > Because if we are talking about IP networking I have two
>         >     observations:
>         >      >
>         >      > A) If you need to traverse via a specific node (ie.
>         firewall) you
>         >     better
>         >      > apply IP encapsulation to that node.. I don't think IP
>         >     encapsulation can
>         >      > be hijacked today such that destination address of
>         the packet is
>         >     ignored.
>         >      >
>         >      > B) Have you seen any IP network where upon topology
>         change (link
>         >     or node
>         >      > failure) you suddenly start dropping flows in spite
>         of SPT offering
>         >      > perhaps few ms longer path with 10 ms more jitter ?
>         >      >
>         >      > Or are some SR marketing slides promise to turn IP
>         networks in
>         >      > something new ? Worse ... do they mention path
>         quality guarantees,
>         >      > resource reservations ? I hope not.
>         >      >
>         >      > Thx,
>         >      > R.
>         >      >
>         >      >
>         >      >
>         >      >
>         >      >
>         >      >
>         >      >
>         >      >
>         >      >
>         >      > On Mon, Aug 3, 2020 at 8:10 PM Joel M. Halpern
>         <[e-mail]
>         <mailto:[e-mail]%0b>>    
>         <mailto:[e-mail]%20%0b>>
>         <mailto:[e-mail]>> wrote:
>         >      >
>         >      > Well less serious for TE SIDs, I am not sure the
>         problem is
>         >     restricted
>         >      > to just service SIDs.
>         >      >
>         >      > Suppose that the PCE has specified the path to meet
>         some complex te
>         >      > objective.  The bypass node has no way of knowing
>         what those
>         >      > constraints
>         >      > were.  And for some kinds of traffic, it is better to
>         drop the packet
>         >      > than to deliver it outside the envelop.  I suspect
>         that the right
>         >      > answer
>         >      > to this is "too bad". If so, as with the distinction
>         regarding
>         >     service
>         >      > nodes, we should say so, shouldn't we?
>         >      >
>         >      > Yours,
>         >      > Joel
>         >      >
>         >      > On 8/3/2020 2:36 AM, Alexander Vainshtein wrote:
>         >      > > Mach, Joel and all,
>         >      > >
>         >      > > I think that in most cases:
>         >      > >
>         >      > > 1.There is clear differentiation between
>         "topological" and
>         >     "service"
>         >      > > instructions in SID advertisements. E.g.:
>         >      > >
>         >      > > oIGP Prefix Node SIDs IGP Adj-SIDs (identified as
>         such in the
>         >      > > corresponding IGP advertisements) represent topological
>         >     instructions
>         >      > >
>         >      > > oService SIDs for SRv6 (see SRv6 BGP-Based Overlay
>         Services
>         >      > >
>         >      >
>         >    
>         <https://clicktime.symantec.com/3CCcy9mY6cMfbk7QfjhiA3R6H2?u=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-ietf-bess-srv6-services-04
>         <https://clicktime.symantec.com/3CCcy9mY6cMfbk7QfjhiA3R6H2?u=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-ietf-bess-srv6-services-04%0b>>    
>         <https://clicktime.symantec.com/3GC5af2z3JzphZDkPncHAQi6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-ietf-bess-srv6-services-04__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo4T-L0nl%24>>
>         >      >
>         >      > > draft) unsurprisingly represent “service” instructions
>         >      > >
>         >      > > 2.Segments that represent topological instructions
>         can be bypassed,
>         >      > > while segments that represent service instructions
>         require
>         >      > alternative
>         >      > > protection mechanisms.
>         >      > >
>         >      > > This view seems to be aligned with RFC 8402
>         >      > >
>         <https://clicktime.symantec.com/345NLCB6TydxuuqUjtcDRZy6H2?u=https%3A%2F%2Ftools.ietf.org%2Fhtml%2Frfc8402
>         <https://clicktime.symantec.com/345NLCB6TydxuuqUjtcDRZy6H2?u=https%3A%2F%2Ftools.ietf.org%2Fhtml%2Frfc8402%0b>>    
>         <https://clicktime.symantec.com/37PzUKAD82cjSvmVcGvpkhF6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Ftools.ietf.org%2Fhtml%2Frfc8402__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo0I4Ybtm%24>>
>         >     that says in Section 1:
>         >      > >
>         >      > >     In the context of an IGP-based distributed
>         control plane, two
>         >      > >
>         >      > > topological segments are defined: the IGP-Adjacency
>         segment and the
>         >      > >
>         >      > >     IGP-Prefix segment.
>         >      > >
>         >      > >     In the context of a BGP-based distributed
>         control plane, two
>         >      > >
>         >      > > topological segments are defined: the BGP peering
>         segment and the
>         >      > >
>         >      > >     BGP-Prefix segment.
>         >      > >
>         >      > > In the case of SR-MPLS this differentiation is
>         assumed in Section
>         >      > 3.4 of
>         >      > > the Node Protection for SR-TE Path
>         >      > >
>         >      >
>         >    
>         <https://clicktime.symantec.com/3JbSWGx5DAfNPsZdhpExxx96H2?u=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-hegde-spring-node-protection-for-sr-te-paths-07%23section-3.4
>         <https://clicktime.symantec.com/3JbSWGx5DAfNPsZdhpExxx96H2?u=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-hegde-spring-node-protection-for-sr-te-paths-07%23section-3.4%0b>>    
>         <https://clicktime.symantec.com/3CrUgARW8somAbw6TisjgJ16H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-hegde-spring-node-protection-for-sr-te-paths-07%2Asection-3.4__%3BIw%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo9wO-Ssn%24>>
>         >      >
>         >      > > draft that says:
>         >      > >
>         >      > >     The node protection mechanism described in the
>         previous
>         >     sections
>         >      > >
>         >      > >     depends on the assumption that the label
>         immediately below
>         >      > the top
>         >      > >
>         >      > > label in the label stack is understood in the IGP
>         domain.  When the
>         >      > >
>         >      > >     provider edge routers exchange service labels
>         via BGP or some
>         >      > other
>         >      > >
>         >      > >     non-IGP mechanism the bottom label is not
>         understood in the IGP
>         >      > >
>         >      > >     domain.
>         >      > >
>         >      > >     The egress node protection mechanisms described
>         in the draft
>         >      > >
>         >      > >     [RFC8679
>         <https://clicktime.symantec.com/3JfvtBAmaQPN1jA3pM6TdCE6H2?u=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc8679
>         <https://clicktime.symantec.com/3JfvtBAmaQPN1jA3pM6TdCE6H2?u=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc8679%0b>>    
>         <https://clicktime.symantec.com/36dMAjuYTQovo8jHwmm3eJw6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc8679__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo8MGipXc%24>>]
>         >     is
>         >      > > applicable to this use case and no additional changes
>         >      > >
>         >      > >     will be required for SR based networks
>         >      > >
>         >      > > The scenarios in which  differentiation between
>         “topological” and
>         >      > > “service” instructions is broken are indeed
>         problematic. E.g.,
>         >      > consider
>         >      > > the use case in which a Node SID in the ERO of a
>         SR-TE path
>         >      > identifies a
>         >      > > node that acts as a firewall for all packets it
>         receives, i.e.,
>         >      > provides
>         >      > > the firewall service without any dedicated service SID
>         >      > identifying it.
>         >      > > One could say that the Node SID of such a node
>         would combine
>         >      > topological
>         >      > > and service instructions thus breaking the
>         differentiation
>         >      > between the two.
>         >      > >
>         >      > > I am not sure if usage of such “combined” SIDs
>         could be prevented
>         >      > or at
>         >      > > least discouraged.
>         >      > >
>         >      > > If not, providing an ability to identify such SIDs
>         in the
>         >      > advertisement
>         >      > > mechanisms would be useful IMHO.
>         >      > >
>         >      > > My 2c,
>         >      > >
>         >      > > Sasha
>         >      > >
>         >      > > Office: [phone]
>         >      > >
>         >      > > Cell: [phone]
>         >      > >
>         >      > > Email: [e-mail]
>         <mailto:[e-mail]>
>         >     <mailto:[e-mail]>
>         >      > <mailto:[e-mail]>
>         >      > >
>         >      > > -----Original Message-----
>         >      > > From: spring <[e-mail]
>         <mailto:[e-mail]%0b>>    
>         <mailto:[e-mail]%0b>>
>         >     <mailto:[e-mail]>> On Behalf Of Mach Chen
>         >      > > Sent: Monday, August 3, 2020 6:30 AM
>         >      > > To: Joel M. Halpern <[e-mail]
>         <mailto:[e-mail]%0b>>    
>         <mailto:[e-mail]%0b>> <mailto:[e-mail]>>;
>         > [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]> <mailto:[e-mail]>
>         >      > > Subject: Re: [spring] Spring protection -
>         determining applicability
>         >      > >
>         >      > > Hi Joel,
>         >      > >
>         >      > > I think this is a good point that may not be
>         discussed in the
>         >      > past. And
>         >      > > I also don't think there is a "can be bypassed"
>         indication in the
>         >      > > routing advertisement for now.
>         >      > >
>         >      > > IMHO, the information advertised by routing is
>         neutral, such
>         >      > information
>         >      > > (can or cannot be bypassed) is more path specific, thus
>         >     normally the
>         >      > > controller should be responsible for deciding
>         whether/which SID
>         >      > can be
>         >      > > bypassed.
>         >      > >
>         >      > > Best regards,
>         >      > >
>         >      > > Mach
>         >      > >
>         >      > >  > -----Original Message-----
>         >      > >
>         >      > >  > From: spring [mailto:[e-mail]
>         <mailto:[e-mail]>
>         >      > <mailto:[e-mail]>
>         >    
>         <mailto:[e-mail]%0b%3e%20%3cmailto:[e-mail]%3e>]
>         >     On Behalf Of Joel M.
>         >      > >
>         >      > >  > Halpern
>         >      > >
>         >      > >  > Sent: Monday, August 3, 2020 7:51 AM
>         >      > >
>         >      > >  > To: [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]>
>         >     <mailto:[e-mail]>
>         >      > <mailto:[e-mail] <mailto:[e-mail]
>         <mailto:[e-mail]%20%3cmailto:[e-mail]%0b>>    
>         <mailto:[e-mail]%20%3cmailto:[e-mail]>>>
>         >      > >
>         >      > >  > Subject: [spring] Spring protection -
>         determining applicability
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > (WG Chair hat Off, this is merely a note from a
>         slightly
>         >      > confused WG
>         >      > >
>         >      > >  > participant.)
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > I have been reading the various repair drafts,
>         and the various
>         >      > >
>         >      > >  > networks programming and service programming
>         draft, and I am
>         >      > trying to
>         >      > >
>         >      > >  > figure out one aspect of the combination.
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > How does a node that is doing some form of
>         bypass (suppose, for
>         >      > >
>         >      > >  > simplicity, it is Node N2 deciding to bypass the
>         next SID for
>         >      > a failed
>         >      > >
>         >      > >  > node N3) know that it is safe to do so?
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > If the path was just for TE, then it is "safe"
>         if the new path
>         >      > meets
>         >      > >
>         >      > >  > the TE criteria.  or maybe it is safe if it is
>         even close, as
>         >      > long as
>         >      > >
>         >      > >  > it is not used for too long.
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > But what if the node were a Firewall, included
>         to meet legal
>         >      > > requirements?
>         >      > >
>         >      > >  > Or was some other necessary programmatic
>         transform (wince we are
>         >      > >
>         >      > >  > deliberately vague about what nodes can do when
>         asked suitably.)
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > Is there some "can be bypassed" indication in
>         the routing
>         >      > >
>         >      > >  > advertisements that I missed?
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > Thank you,
>         >      > >
>         >      > >  > Yours,
>         >      > >
>         >      > >  > Joel
>         >      > >
>         >      > >  >
>         >      > >
>         >      > >  > _______________________________________________
>         >      > >
>         >      > >  > spring mailing list
>         >      > >
>         >      > >  > [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]>
>         >     <mailto:[e-mail]>
>         >      > <mailto:[e-mail] <mailto:[e-mail]
>         <mailto:[e-mail]%20%3cmailto:[e-mail]%0b>>    
>         <mailto:[e-mail]%20%3cmailto:[e-mail]>>>
>         >      > >
>         >      > >  >
>         >      > >
>         >      >
>         >
>         https://clicktime.symantec.com/367qhU4KiUkzW9uGC4eAvP46H2?u=https%3A%2
>         <https://clicktime.symantec.com/367qhU4KiUkzW9uGC4eAvP46H2?u=https%3A%252>
>         >    
>         <https://clicktime.symantec.com/3Q7vX2qWSUdWVc892XuXy2H6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fclicktime.symantec.com%2F367qhU4KiUkzW9uGC4eAvP46H2%3Fu%3Dhttps%2A3A%2A2__%3BJSU%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo6HwPLil%24>
>         >      > >
>         >      >
>         >    
>         <https://clicktime.symantec.com/367qhU4KiUkzW9uGC4eAvP46H2?u=https%3A%252
>         <https://clicktime.symantec.com/367qhU4KiUkzW9uGC4eAvP46H2?u=https%3A%252%0b>>    
>         <https://clicktime.symantec.com/3A5B8H2Fm1rPnaZ3Supjwr6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fclicktime.symantec.com%2F367qhU4KiUkzW9uGC4eAvP46H2%3Fu%3Dhttps%2A3A%2A252__%3BJSU%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDozoQiAHk%24>>
>         >      > >
>         >      > >  > F%2Fwww.ietf.org <http://2Fwww.ietf.org>
>         >    
>         <https://clicktime.symantec.com/39NznmYBtRuHARhGJ5W5dGB6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__http%3A%2F2Fwww.ietf.org__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo-pPCjvR%24>
>         >      >
>         <https://clicktime.symantec.com/3GWT9fyjaFi3FcvHDvoodvS6H2?u=http%3A%2F%2F2Fwww.ietf.org
>         <https://clicktime.symantec.com/3GWT9fyjaFi3FcvHDvoodvS6H2?u=http%3A%2F%2F2Fwww.ietf.org%0b>>    
>         <https://clicktime.symantec.com/39NznmYBtRuHARhGJ5W5dGB6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__http%3A%2F2Fwww.ietf.org__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo-pPCjvR%24>>%2Fmailman%2Flistinfo%2Fspring
>         >      > >
>         >      > > _______________________________________________
>         >      > >
>         >      > > spring mailing list
>         >      > >
>         >      > > [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]>
>         >     <mailto:[e-mail]> <mailto:[e-mail]
>         <mailto:[e-mail]%0b>>     <mailto:[e-mail]%0b>>
>         <mailto:[e-mail]>>
>         >      > >
>         >      > >
>         >      >
>         >
>         https://clicktime.symantec.com/367qhU4KiUkzW9uGC4eAvP46H2?u=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring
>         >    
>         <https://clicktime.symantec.com/3BhyEtx4Q7n74BhiRnfMJtT6H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fclicktime.symantec.com%2F367qhU4KiUkzW9uGC4eAvP46H2%3Fu%3Dhttps%2A3A%2A2F%2A2Fwww.ietf.org%2A2Fmailman%2A2Flistinfo%2A2Fspring__%3BJSUlJSUl%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo-hR3gAD%24>
>         >      > >
>         >      > >
>         >      > >
>         >      > >
>         >      >
>         >
>         ------------------------------------------------------------------------
>         >      > > Notice: This e-mail together with any attachments
>         may contain
>         >      > > information of Ribbon Communications Inc. that is
>         confidential
>         >      > and/or
>         >      > > proprietary for the sole use of the intended
>         recipient. Any review,
>         >      > > disclosure, reliance or distribution by others or
>         forwarding
>         >     without
>         >      > > express permission is strictly prohibited. If you
>         are not the
>         >      > intended
>         >      > > recipient, please notify the sender immediately and
>         then delete all
>         >      > > copies, including any attachments.
>         >      > >
>         >      >
>         >
>         ------------------------------------------------------------------------
>         >      > >
>         >      > > _______________________________________________
>         >      > > spring mailing list
>         >      > > [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]> <mailto:[e-mail]>
>         >      > >
>         https://clicktime.symantec.com/3Q1xsKGyMzNJCKyVPByhqLq6H2?u=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring
>         >    
>         <https://clicktime.symantec.com/3NrDnSTReXh671G79BVGEq16H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo5KlPnbj%24>
>         >      > >
>         >      >
>         >      > _______________________________________________
>         >      > spring mailing list
>         >      > [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]> <mailto:[e-mail]>
>         >      >
>         https://clicktime.symantec.com/3Q1xsKGyMzNJCKyVPByhqLq6H2?u=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring
>         >    
>         <https://clicktime.symantec.com/3NrDnSTReXh671G79BVGEq16H2?u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqguoZHgwp6vpRHOGt8AkTRDuiDo5KlPnbj%24>
>         >      >
>         >
>         > _______________________________________________
>         >     spring mailing list
>         > [e-mail] <mailto:[e-mail]>
>         <mailto:[e-mail]>
>         >
>         https://clicktime.symantec.com/3Q1xsKGyMzNJCKyVPByhqLq6H2?u=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring
>         >
>         >
>         <https://clicktime.symantec.com/3NrDnSTReXh671G79BVGEq16H2?u=https%3A%
>         <https://clicktime.symantec.com/3NrDnSTReXh671G79BVGEq16H2?u=https%3A%25%0b>>
>         2F%2Furldefense.com%2Fv3%2F__https%3A%2Fwww.ietf.org
>         <http://2Fwww.ietf.org>%2Fmailman%2Flisti
>         >
>         nfo%2Fspring__%3B%21%21NEt6yMaO-gk%21S0Yusx9FYNE8E_R1oiQAtQxgm0x0wxqgu
>         > oZHgwp6vpRHOGt8AkTRDuiDo5KlPnbj%24>
>         >
>         >
>         >
>         >
>         ----------------------------------------------------------------------
>         > --
>         > Notice: This e-mail together with any attachments may contain
>         > information of Ribbon Communications Inc. that is
>         confidential and/or
>         > proprietary for the sole use of the intended recipient. Any
>         review,
>         > disclosure, reliance or distribution by others or forwarding
>         without
>         > express permission is strictly prohibited. If you are not
>         the intended
>         > recipient, please notify the sender immediately and then
>         delete all
>         > copies, including any attachments.
>         >
>         ----------------------------------------------------------------------
>         > --
>
>         _______________________________________________
>         spring mailing list
>         [e-mail] <mailto:[e-mail]>
>         https://clicktime.symantec.com/3Q1xsKGyMzNJCKyVPByhqLq6H2?u=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring
>         _______________________________________________
>         spring mailing list
>         [e-mail] <mailto:[e-mail]>
>         https://clicktime.symantec.com/3Q1xsKGyMzNJCKyVPByhqLq6H2?u=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring
>
>         ------------------------------------------------------------------------
>
>         Notice: This e-mail together with any attachments may contain
>         information of Ribbon Communications Inc. that is confidential
>         and/or proprietary for the sole use of the intended recipient.
>         Any review, disclosure, reliance or distribution by others or
>         forwarding without express permission is strictly prohibited.
>         If you are not the intended recipient, please notify the
>         sender immediately and then delete all copies, including any
>         attachments.
>
>         ------------------------------------------------------------------------
>
>
> _______________________________________________
> spring mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/spring