[manet] Opsdir early review of draft-ietf-manet-dlep-credit-fl ow-control-15
Mohamed Boucadair via Datatracker <[email protected]>
| Newsgroups | gmane.ietf.manet |
|---|---|
| Message-ID | <172182564578.727884.14028482688083341345@dt-datatracker-659f84ff76-9wqgv> |
Reviewer: Mohamed Boucadair Review result: Has Issues Hi all, Thanks for the authors for the effort put into this well-written document. First, I'm always impressed by authors who persevere and push for a specification over many years. I checked the archives to see whether there were contentious points that would justify the long development process, but I didn't find something meaningful other than some discussions on how the various pieces (this I-D, classification, etc.) can be documented. That discussion answered a comment I had about the need to separate the classification spec from this. I won't thus raise that point in my review. However, it seems weird (at least to me) to define objects and put in the abstract that future documents "will mandate the use". I was also expecting to see a discussion on how this flow-control is superior compared to the pause approach specified in RFC 8651, including a discussion about co-existence considerations and which one will take precedence. Overall, the document (seems) to reason following the model in Figure 1 of RFC 8175, while it should be applicable as well for the configuration in Figure 2 of 8175. For example, the FID uniqueness (at the router side) should be associated with the link over which the packets will be sent. >From a protocol machinery standpoint, there are some few cases where I think the MUST behavior is not justified. Please refer to the link below for more details on this. >From an ops standpoint, the document includes a dedicated section on management. However, I think that more concrete implementation behavior should be provided, e.g., * how to report errors? * expose configuration knobs to control many of the parameters there. Also, exposing implementation default would be helpful when operating the system. * technically characterize some events (e.g., transient events) and provide a minimum value for how frequent messages can be sent. More detailed comments can be found at: https://github.com/boucadair/IETF-Drafts-Reviews/blob/master/2024/draft-ietf-manet-dlep-credit-flow-control-15-rev%20Med.pdf hope this helps Cheers, Med _______________________________________________ manet mailing list -- [email protected] To unsubscribe send an email to [email protected]