Re: [Lsr] [rt5.ietf.org #7080] System ID in ISIS

Hi Les,

Thanks for the response.
Yes, i understand that it would require a lot of efforts and can take some
years.

But as we discussed SYS ID 0 should  be considered valid as Standard
Documents doesn't define otherwise and at same time we should try to use a
logical value as a SYS ID so we don't create inter-operability issues where
a vendor doesn't consider it as valid.

"What reason did the customer give for configuring a systemid of 0?"

Customer had only a few Cisco nodes participating in the IS-IS , so they
started configuring the sys id 0000.0000.0000 then 0000.0000.0001 and so on.

Thanks a lot for your time and understanding.

Regards
Jaideep

On Thu, 16 Jun, 2022, 12:57 am Les Ginsberg (ginsberg), <ginsberg@cisco.com>
wrote:

> Jaideep –
>
>
>
> I am not unsympathetic to problems encountered in the field.
>
> It isn’t always easy to get a customer to agree to what you and I might
> easily agree makes sense.
>
>
>
> But, in this case, consider what might be required to get an unambiguous
> standard defined:
>
>
>
> 1)We would have to establish consensus on whether 0 should/should not be
> considered as valid
>
>
>
> 2)We would have to get the vendors whose implementations do not conform to
> whatever is agreed upon in #1 to commit to changing their implementations
>
>
>
> 3)A standard would have to be written and work its way through the
> approval process. This would most practically be an IETF draft as there is
> no one actively updating ISO specs.
>
>
>
> All of this would take years – and at the end we would only have resolved
> the use of a systemid for which there is no practical deployment case.
>
>
>
> I just don’t think this is worth the effort.
>
>
>
> What reason did the customer give for configuring a systemid of 0?
>
>
>
>    Les
>
>
>
> *From:* Jaideep Choudhary <jaideepchoudhary91@gmail.com>
> *Sent:* Tuesday, June 14, 2022 11:59 PM
> *To:* Les Ginsberg (ginsberg) <ginsberg@cisco.com>
> *Subject:* Re: [Lsr] [rt5.ietf.org #7080] System ID in ISIS
>
>
>
> Hi Les,
>
>
>
> We have recently experienced some issue in this regards on a vendor where
> SYS ID of 0000.0000.0000 was not interpreted correctly.
>
>
>
> In a single area with level L1 and mutli-vendor setup, a Cisco node was
> configured with SYS I'd of 0000.0000.0000.
>
> The other vendor node could not interpret 0000.0000.0000 as a system id
> the way it does for other SYS IDs and there were some issues with the
> system level command outputs.
>
>
>
> Modifying the SYS id  on Cisco node did help, but it took a lot of time to
> find the cause.
>
>
>
> When I talked about routing issues, what I meant was, that if for example
> a Juniper node doesn't consider SYS ID of 0000.0000.0000 as legal, then it
> may not install the LSP from the node with SYS ID=0.
>
>
>
> If it is directly connected node, then Juniper node may not form adjacency
> and it can be found out easily, but if it is not directly connected node,
> then it would take good amount of time to find the cause.
>
>
>
> That in turn can cause some issues.
>
>
>
> I also agree 100% that it doesn't make sense to make a SYS ID of 0, but
> talking from an experience we had, it was configured.
>
>
>
> Since it is not defined as invalid as per standard documents, it also
> makes sense to have uniformity across different implementations, so no such
> issue occurs in multi-vendor setups.
>
>
>
> Also the reason of verifying this with IETF was to understand the reason
> behind Juniper defining sys I'd as illegal.
>
> We wanted to confirm if SYS ID 0 is reserved for some other use ?
>
>
>
> I hope , I am able to make my point here.
>
>
>
> Really appreciate your time.
>
> Thanks !
>
>
>
> Regards
>
> Jaideep
>
>
>
>
>
> On Wed, 15 Jun, 2022, 10:46 am Les Ginsberg (ginsberg), <
> ginsberg@cisco.com> wrote:
>
> (Taking this offlist – BCC the WG)
>
>
>
> Jaideep –
>
>
>
> From a standards perspective I have provided you with what I know.
>
>
>
> To characterize this as something which can cause “serious routing issues”
> is an exaggeration.
>
> Given that the same system ID cannot be used on more than one router, at
> worst if you were in a deployment where an implementation did not accept a
> systemid of 0000, all you would need to do is modify the config of a single
> router.
>
>
>
> Assigning a systemid which has no relationship to the identity of the
> equipment/configuration of the node isn’t practical – I don’t think any
> thoughtful network manager would ever do such a thing.
>
> In my view you have lost perspective on this issue.
>
>
>
>    Les
>
>
>
>
>
> *From:* Jaideep Choudhary <jaideepchoudhary91@gmail.com>
> *Sent:* Tuesday, June 14, 2022 9:57 PM
> *To:* Les Ginsberg (ginsberg) <ginsberg@cisco.com>
> *Cc:* Tony Li <tony.li@tony.li>; support@ietf.org; lsr@ietf.org
> *Subject:* Re: [Lsr] [rt5.ietf.org #7080] System ID in ISIS
>
>
>
> Hi Les,
>
>
>
> Thanks for the quick response.
>
>
> I also could not find anywhere in the standard documentation stating that
> SYS ID of 0000.0000.0000 in IS-IS as invalid nor is there any restriction
> to how to calculate the SYS ID.
>
>
>
> Yes, there are recommendations to use MAC or IP address to calculate the
> SYS ID , so it remains unique in a routing domain, but *couldn't *be
> found anywhere in the standard documentation, if SYS ID *must be derived
> from these addresses only*.
>
> Having said that, in most of the cases, there would be very less
> probability of SYS ID of 0000.0000.0000 being configured in a production
> environment (as you also mentioned), but still, as there is no such
> explicit restriction (in the standards ISO10589 or RFC 3784) to not to use
> SYS ID: 0, so it can still be used as a valid SYS ID in the devices where
> it is allowed to configure the NET/SYSTEM ID manually.
>
>
>
> So in that case if some device the setting of SYS ID being 0 is considered
> as invalid or illegal, that can cause some serious routing issues in a
> single area multi vendor setup in ISIS.
>
> *So, can we say that from Standards perspective SYS ID: 0000.0000.0000 is
> a legal setting ?*
>
>
>
> Regards
>
> Jaideep
>
>
>
> On Tue, Jun 14, 2022 at 9:59 PM Les Ginsberg (ginsberg) <
> ginsberg@cisco.com> wrote:
>
> Jaideep –
>
>
>
> I am not aware that any standard formally defines a system-id of
> 0000.0000.0000 as invalid.
>
> If there is, it would be an ISO specification – but a perusal of ISO
> 10589, ISO 8348, and ISO 7498 did not yield any such statement.
>
> (I would be happy to be corrected if someone has a reference.)
>
>
>
> From a practical standpoint, the lack of agreement on this by all
> implementations should not represent a significant concern.
>
> Schemes which automatically populate the system-id are typically based on
> the MAC address of some NIC on the box.
>
> Another common strategy is to use the zero filled IP address of some
> loopback.
>
> In either case all zeros will not be the result.
>
>
>
> In cases where the systemid is explicitly configured, it is easy enough
> NOT to use all 0’s.
>
>
>
> HTH
>
>
>
>     Les
>
>
>
> *From:* Lsr <lsr-bounces@ietf.org> *On Behalf Of *Jaideep Choudhary
> *Sent:* Tuesday, June 14, 2022 8:00 AM
> *To:* Tony Li <tony.li@tony.li>
> *Cc:* support@ietf.org; lsr@ietf.org
> *Subject:* Re: [Lsr] [rt5.ietf.org #7080] System ID in ISIS
>
>
>
> Hi Tony,
>
>
>
> I am not looking for technical support, but looking for IETF's perspective
> regarding the system id in IS-IS.
>
>
>
> As per the RFC 3784 there is no mention about any invalid value in a
> system id.
>
>
>
> Can you please confirm whether there is any such restriction to not to use
> a SYS ID of 0000.0000.0000 as per IETF standards ?
>
>
>
> If this mailing address is not appropriate for answering this query, can
> you suggest/redirect me to the correct team from IETF ?
>
>
>
> Thanks.
>
>
>
> Regards
>
> Jaideep
>
>
>
> On Tue, Jun 14, 2022, 20:19 Tony Li <tony.li@tony.li> wrote:
>
>
>
> Hi,
>
>
>
> Neither of these mailing lists are appropriate for technical support.
> Please contact your vendors directly.
>
>
>
> Tony
>
>
>
>
>
> On Jun 14, 2022, at 12:12 AM, Jaideep Choudhary <
> jaideepchoudhary91@gmail.com> wrote:
>
>
>
> Hi Team,
>
> I would like to know, whether in IS-IS, a system id can be 0000.0000.0000
> or it is an invalid value for sys I'd ?
>
> As per ISO 10589 a system id can be of 1 to 8 bytes long, but doesn't
> mention explicitly whether SYS ID of 0000.0000.0000 could be invalid.
>
> Also as per RFC 3784, it says System id is typically of 6 bytes, but
> doesn't talk about any invalid option.
>
> The reason I am asking this is that Juniper defines a SYS ID of
> 0000.0000.0000 as invalid.
>
>
> https://www.juniper.net/documentation/us/en/software/junos/is-is/topics/concept/is-is-routing-overview.html
>
>
>
> This can cause issues in inter-operability as some vendors like Cisco
> doesn't define a SYS-ID of 0000.0000.0000 as invalid.
>
> I would appreciate your response on this.
>
> Regards
>
> Jaideep Choudhary
>
>
>
> On Mon, 13 Jun, 2022, 11:08 pm Cindy Morgan via RT, <support@ietf.org>
> wrote:
>
> Hi Jaideep,
>
> You have reached the IETF Secretariat, which is the administrative branch
> of the IETF, and as such, we are not qualified to answer your technical
> questions.
>
> You might have better luck if you try posing your question to the Link
> State Routing (LSR) Working Group (
> https://datatracker.ietf.org/wg/lsr/about/). LSR was formed by merging
> the ISIS and OSPF WGs and assigning all their existing adopted work at the
> time of chartering to LSR. Their mailing list address is lsr@ietf.org.
>
> Best regards,
> Cindy
>
> On Mon Jun 13 10:10:54 2022, jaideepchoudhary91@gmail.com wrote:
>
> Hi Team,
>
>
>
> I would like to know, whether in IS-IS, a system id can be 0000.0000.0000
> or it is an invalid value for sys I'd ?
>
>
>
> As per ISO 10589 a system id can be of 1 to 8 bytes long, but doesn't
> mention explicitly whether SYS ID of 0000.0000.0000 could be invalid.
>
>
>
> Also as per RFC 3784, it says System id is typically of 6 bytes, but
> doesn't talk about any invalid option.
>
>
>
> The reason I am asking this is that Juniper defines a SYS ID of
> 0000.0000.0000 as invalid.
>
>
>
>
>
>
> https://www.juniper.net/documentation/us/en/software/junos/is-is/topics/concept/is-is-routing-overview.html
>
>
>
> This can cause issues in inter-operability as some vendors like Cisco
> doesn't define a SYS-ID of 0000.0000.0000 as invalid.
>
>
>
> I would appreciate your response on this.
>
>
>
> Regards
>
> Jaideep Choudhary
>
>
>
>
>
>
>
> _______________________________________________
> Lsr mailing list
> Lsr@ietf.org
> https://www.ietf.org/mailman/listinfo/lsr
>
>
>
>