Speaking as WG member: I agree with Les. A non-backward compatible change is a non-starter.
I’m not sure why you’d need to present this again at the Interim unless you provide backward compatibility. Thanks, Acee From: Lsr <lsr-boun...@ietf.org> on behalf of "Les Ginsberg (ginsberg)" <ginsberg=40cisco....@dmarc.ietf.org> Date: Friday, August 12, 2022 at 12:05 PM To: 龚立艳 <gongli...@chinamobile.com>, "lsr@ietf.org" <lsr@ietf.org>, shraddha <shrad...@juniper.net> Subject: Re: [Lsr] Comments ondraft-gong-lsr-exclusive-link-for-flex-algo Liyan – You agree that there is an existing way to prune links from the IGP SPF. Still, you insist that an extension that requires ALL routers – whether they support flex algo or not – to utilize a new advertisement when computing algo 0 SPF is necessary. Your rationale for this is that this allows you to use IGP Metric for flex algo in cases where the IGP metric would have been set to maximum. But we already have the ability to define metrics specific to flex algo – and this is greatly enhanced by the generic metric defined in https://datatracker.ietf.org/doc/draft-ietf-lsr-flex-algo-bw-con/ Requiring a non backwards compatible extension to be used in base protocol operation in order to support a new feature is exactly what we MUST NOT do when introducing protocol extensions. My opinion is unchanged – this is a bad idea – and completely unnecessary. Les From: 龚立艳 <gongli...@chinamobile.com> Sent: Friday, August 12, 2022 2:16 AM To: lsr@ietf.org; Les Ginsberg (ginsberg) <ginsb...@cisco.com>; shraddha <shrad...@juniper.net> Subject: Re:Re: [Lsr] Comments ondraft-gong-lsr-exclusive-link-for-flex-algo Hi Shraddha and Les, Sorry for late reply and thanks for your comments. Yes, the maximum link metric mechanism is an option as described in section 4 of the draft. But, it has two defects which we also wanted to discuss in ietf 114 meeting. Firstly, it restricts the Flex-Algorithm from using IGP-Cost as its metric-type. Secondly, it does not work with OSPF. For OSPF,the links with maximummetric value(65535) are also included in the SPF calculation,even if not preferred. If there are no other preferred paths,the Flex-Algoritnm links will still affect the result of thenormal SPF calculation. Due to the time constraints,The presentation has been moved to the interim meeting on 2022-09-21. For more detail, please refer to the slides. https://datatracker.ietf.org/meeting/114/materials/slides-114-lsr-13-exclusive-link-for-flex-algo-00. In view of these two cases, new protocol extension becomes necessary. As for the backward incompatible issues, we think it can be avoided by deployment. For example, the new extension should be deployed in sync with the Flex-Algo feature, so that all the routers in one IGP domain will run the same software version. Looking forward to your reply. Best Regards, Liyan ----邮件原文---- 发件人："Les Ginsberg \\(ginsberg\\)<file:///(ginsberg/)>" <ginsberg=40cisco....@dmarc.ietf.org<mailto:ginsberg=40cisco....@dmarc.ietf.org>> 收件人：Shraddha Hegde <shraddha=40juniper....@dmarc.ietf.org<mailto:shraddha=40juniper....@dmarc.ietf.org>>,"lsr@ietf.org<mailto:lsr@ietf.org>" <lsr@ietf.org<mailto:lsr@ietf.org>> 抄 送: (无) 发送时间：2022-07-29 21:14:08 主题：Re: [Lsr] Comments on draft-gong-lsr-exclusive-link-for-flex-algo I fully agree with Shraddha. In fact Section 4 of the draft makes clear why no protocol extensions are needed. Les From: Lsr <lsr-boun...@ietf.org<mailto:lsr-boun...@ietf.org>> On Behalf Of Shraddha Hegde Sent: Friday, July 29, 2022 2:18 AM To: lsr@ietf.org<mailto:lsr@ietf.org> Subject: [Lsr] Comments on draft-gong-lsr-exclusive-link-for-flex-algo Authors, I suggest that the usecase can be satisfied using the backward compatible Maximum link metric mechanism defined in the draft. I don’t see any need to define protocol extensions, that are backward incompatible and can cause serious issues in the network in the presence of older implementations. Rgds Shraddha Juniper Business Use Only
_______________________________________________ Lsr mailing list Lsr@ietf.org https://www.ietf.org/mailman/listinfo/lsr
