Re: [Isis-wg] AD review of draft-ietf-isis-mi-bis-01

[Alexander Okonnikov <[e-mail]>]

Les,

Exactly, I examined 10589, and there medium-dependent aspects are mentioned using "subnetwork" term or "link". "Subnetwork-dependent" and "subnetwork-independent", "broadcast subnetworks", "general topology subnetworks", where "point-to-point links" are one of the latter. On the other hand, term "circuit" is mentioned as "circuit metric/cost", "circuit ID", "circuit database", "circuit type broadcast/p2p", "adjacency" and so on. For example, circuit database can be filled in with "broadcast" and "p2p" circuits, but IS-IS doesn't care whether latter is "p2p" or "p2p over lan". It will use the same procedures for both of them (establishment of adjacency, database synchronization). In other words, I would want to say that term "circuit" is overloaded here - while in 10589 "circuit" means entity of the protocol, in MI-ISIS it means what 10589 calls "subnetwork/link".

Regarding AllIS - yes, some implementations use AllIS only for IIH, and AllL1IS/AllL2IS for SNP/LSP. While it is not beneficial to use AllIS for SNP and LSP PDUs, other implementations behave in this manner, and this is valid behavior.

You have specified that LSP and SNP for standard instance will use AllL1IS/AllL2IS. It is not clear whether AllIS is allowed for MI-capable router for using it in standard instance over P2PoverLAN circuits for SNP and LSP, as it is allowed for non-MI-capable router (per RFC 5309).

Thank you.

On 17 марта 2017 г., 22:55 +0300, Les Ginsberg (ginsberg) <[e-mail]>, wrote:
> Alexander –
>
>
> From: Alexander Okonnikov [mailto:[e-mail]]
> Sent: Friday, March 17, 2017 11:40 AM
> To: Alia Atlas; Les Ginsberg (ginsberg)
> Cc: [e-mail]; [e-mail]
> Subject: Re: [Isis-wg] AD review of draft-ietf-isis-mi-bis-01
>
> Hi Alia and authors,
>
> Regarding point 2): This was root of my confusion and reason for submitting Errata ID #4521. Per my understanding a IS-IS circuit is a representaion of layer 2 link for IS-IS, and not a physical circuit itself. For mentioning type of physical media ISO 10589:2002 has another term subnetwork. Early versions of draft-ietf-isis-mi used term "network" (up to -04), which was replaced by "circuit" later. In my mind, term "circuit" creates ambiguity. From IS-IS point of view regular P2P circuit and P2PoverLAN circuit are both P2P circuits. May be it is better to create dedicated section for P2PoverLAN "circuits" to avoid confusion.
>
> [Les:] The word “circuit” was chosen precisely because it is the term used in ISO 10589. You could start with ISO 10589 Section 3.6.3 – and then examine the hundreds of other uses of the term in the document.
>
> Errata ID #4520 was not incorporated fully. Info regarding AllIS MAC was added, but -bis still clarifies usage of standard MACs (for standard instance) and new ones (for non-zero instances) only for IIHs in P2PoverLAN mode. I.e. the draft doesn't specify which MACs should be used for SNPs and LSPs in that mode. From current text we can conclude that for standard instance addresses AllL1IS and AllL2IS could be used on P2PoverLAN. But AllIS could be used as well, and not only for IIH. That is why the errata mentions "PDU", rather than "IIH".
>
> [Les:] RFC 5309 is relaxed in its language on this point. It says (Section 4.1):
>
> “The Multi-destination address including AllISs, AllL1ISs, and
>    AllL2ISs, defined in [ISO10589], can be used for link-layer
>    encapsulation; the use of AllISs is recommended.”
>
> I have always interpreted this to mean use AllIS for hellos – because Pt2Pt IIHs are not level specific – but use the L1/L2 addresses for SNPs/LSPs which are level specific.
> However, a strict interpretation of RFC 5309 allows that AllISs could be used for SNPs/LSPs. I will correct that omission.
>
> However, we did add text specifying the PDU types explicitly, so I believe any confusion on that point has been addressed.
>
>    Les
>
> Thank you.
>
> Alexander
>
> On 17 марта 2017 г., 21:39 +0300, Les Ginsberg (ginsberg) <[e-mail]>, wrote:
>
> Alia  -
>
> From: Alia Atlas [mailto:[e-mail]]
> Sent: Friday, March 17, 2017 11:13 AM
> To: Les Ginsberg (ginsberg)
> Cc: [e-mail]; [e-mail]
> Subject: Re: AD review of draft-ietf-isis-mi-bis-01
>
> Hi Les,
>
> On Fri, Mar 17, 2017 at 1:57 PM, Les Ginsberg (ginsberg) <[e-mail]> wrote:
> Alia –
>
> Thanx for meticulous review.
> Inline.
>
> From: Alia Atlas [mailto:[e-mail]]
> Sent: Friday, March 17, 2017 8:47 AM
> To: [e-mail]; [e-mail]
> Subject: AD review of draft-ietf-isis-mi-bis-01
>
> As is customary, I have done my AD review of draft-ietf-isis-mi-bis-01.  First, I would like to thank the authors - Les, Stefano, and Wim - for their work on this document.
>
> I do have a few minor issues that need to be fixed ASAP.   I will, however, go ahead and issue a 3 week IETF Last Call for it so that, if the authors act promptly, it can be on the April 13 telechat.
>
>
> 1) This draft pretty clearly should obsolete RFC 6822, but the header claims to merely
> update it.
>
> [Les:] OK
>
>  2) Sec 2.6.2: " Point-to-point adjacency setup MUST be done through the use of the
>          three-way handshaking procedure as defined in [RFC5303] in order to
>          prevent a non-MI capable neighbor from bringing up an adjacency
>          prematurely based on reception of an IIH with an IID-TLV for a non-
>          zero instance."
>      While obviously the 3-way handshake in RFC5303 is fine, given that different MAC addresses are used, I don't see how a non-MI capable neighbor would receive and process an IIH with an IID-TLV where IID != 0.
>
> [Les:] Section 2.6.2 is only talking about true pt-pt media – hence MAC multicast addresses are not relevant. In Section 2.6.1 we discuss the case of operating in Pt-Pt mode on a broadcast circuit:
>
> “When operating in point-to-point mode on a broadcast circuit
>    [RFC5309]…”
>
> Thanks - you are completely correct.
>
>
> >  3) In Appendix A:" Clarification that the IID-TLV is only included in Pt-Pt IIHs
> >    associated with non-zero instances has been added.  This addresses
> >    Errata ID #4519."
> >    In Sec 2.6.2, it says "The presence or absence of the IID-TLV in an IIH indicates that the
> >    neighbor does or does not support this extension, respectively.
> >    Therefore, all IIHs sent on a point-to-point circuit by an MI-RTR
> >    MUST include an IID-TLV.  This includes IIHs associated with IID #0." which is a direct
> >    contradiction and needed since the IID-TLV is used as a capability indication.  The
> >    simplest solution is to remove the claim from the Appendix.
> > [Les:] Again, you are confusing Pt-Pt media with Pt-Pt operation on a Broadcast media. Errata 4519 is concerned with Pt-Pt operation on a LAN – and some modest clarifications in Section 2.6.1 were made to more completely cover the latter case.
> > In the Pt-Pt operation on a LAN case it IS illegal to send IID-TLV in hellos for the zero instance – which Section 2.6.1 explicitly states.
> >
> > Perhaps the Appendix should say:
> >
> > “Clarification that when operating in point-to-point mode on a broadcast circuit the IID-TLV is only included in Pt-Pt IIHs
> >    associated with non-zero instances has been added. ..”
> >
> > Will this address your concern?
>
> Yes, I didn't go and look up Errata ID #4519 to catch the context of pt-to-pt on broadcast only.   That would make it more of a nit than a minor issue, of course.
>
> This all does imply that:
>    a) IIH with IID=0 is a capability advertisement for pt-to-pt links but not pt-to-pt on broadcast.
>    b) It is never ok (discarded by receiver) to send an IIH with IID=0 on a pt-to-pt on broadcast.
>    c) The capability indicator for being an MI-RTR on a pt-to-pt on broadcast link is whether or not the ALLL1MI-IS or ALLL2MI-IS is listened to or sent on.
>
> I know there is real code and deployments behind this - but that's a bit confusing for trouble-shooting & it would be nice to have it more clearly articulated.
> [Les:] From Section 2.6.1:
>
>   MI-RTRs MUST discard IS-IS PDUs received if either of the following
>    is true:
>
>    o  The destination multicast address is AllL1IS, AllL2IS, or AllIS
>       and the PDU contains an IID-TLV
>
>    o  The destination multicast address is one of the two new addresses
>       and the PDU contains an IID-TLV with a zero value for the IID or
>       has no IID-TLV.
>
> From Section 2.6.2:
>
>    The presence or absence of the IID-TLV in an IIH indicates that the
>    neighbor does or does not support this extension, respectively.
>    Therefore, all IIHs sent on a point-to-point circuit by an MI-RTR
>      MUST include an IID-TLV.  This includes IIHs associated with IID #0.
>
> [Les:] I think this language is very precise.
> It is true that operation in Point-to-point mode on a broadcast media is subtly different than operation on true P2P media – which is why we discuss the two in different sections.
> ??
>
> > 4) In the IANA considerations, the references should be updated to point to this document, and most particularly if it obsoletes RFC 6822.
> >
> > [Les:] Agreed. I did not think I had to state that – I assumed IANA would just do that – but I am happy to add a line in IANA section stating that all references to RFC 6822 should be updated to point to this document.
>
> Yes - we need to do that.
>
>
> Incidentally, on another reread (checking the above issues) is there a reason that
> "The destination multicast address is one of the two new addresses
>  and the PDU contains an IID-TLV with a zero value for the IID or
>  has no IID-TLV."
>
> doesn't simply name the addresses (ALLL1MI-IS or ALLL2MI-IS)?  I think that'd be clearer - but this is definitely nit.
>
>
> [Les:] Happy to explicitly call out the specific addresses in the second bullet.
>
>    Les
>
> > Let me know if the above changes will suffice and I will spin a new version.
>
> Sounds good.
>
> Thanks,
> Alia
>
>
> >    Les
> >
> > Regards,
> > Alia
>
> _______________________________________________
> Isis-wg mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/isis-wg