Hi, Bill, Thanks for the review and the note!
Please see inline. > On Jul 13, 2024, at 11:51 AM, Bill Fenner <[email protected]> wrote: > > Hi Carlos, > > My high level question is: why does this belong in ICMP? More particularly, > why would a provider be interested in the sustainability aspects of a single > path as seen by traceroute, as opposed to a wholistic view of the network > that could be gathered by a centralized NMS using a netconf model? Thanks for this high-level question, as we have thought about it. My perspective is that there is no “as opposed to” dichotomy, and instead the wholistic view as well as the on-demand path-specific view are both useful, for different purposes and use cases. I thought we speak to this, but will clarify further. The key, of course, is that the values reported in both cases are consistent. > > Some lower-level feedback: > 1. As described by > https://datatracker.ietf.org/doc/draft-fenner-intarea-probe-clarification/ > due to implementations adding arbitrary data, RFC8335 only permits the single > extension object for extended echo, so this object can't be added to extended > echo at all. Ack. > 2. The Node Throughput Sub-Object is a 32-bit unsigned word in units of bits > per second. That's too small to represent most interface speeds, let alone > aggregate throughput for an entire node. A router with 100 100Gbps > interfaces could have a node throughput over 10Tbps, right, so it seems > prudent to be able to represent more than 4Gbps in this field. (Ditto for > the Component-level Throughput Sub-Object, of course.) Ack as well! > 3. How do UUIDs get mapped to component names? Your example output shows > Present Power(Node=160W,Fan=7W) > Idle Power(Node=152W,Fan=7W,Chassis=10W) > but the TLV for present/idle power contains a uuid. So would the output > really be > Present > Power(e98f3c72-d00c-4690-b9fe-c23b9b19c7e6=160W,15f4db5c-39a6-4936-b566-5901543e43c5=7W) > Idle > Power(e98f3c72-d00c-4690-b9fe-c23b9b19c7e6=152W,15f4db5c-39a6-4936-b566-5901543e43c5=7W,d8574348-11c8-4c25-b36e-75b57d1674f7=10W) > or is there some registry or other mapping mechanism to turn a UUID into a > human readable name? This is a great point. I wonder if a node-level would be a simply identified by the ip address and, say, a zero uuid, perhaps? Otherwise yes, sub-elements would print as you show (lacking mapping), and we will clarify. Thanks! Carlos. > > Thanks, > Bill > > > On Sat, Jul 6, 2024 at 7:39 AM Carlos Pignataro <[email protected] > <mailto:[email protected]>> wrote: >> Hi, Int-Area, >> >> We had posted this draft earlier this year, and have been through some >> review cycles. >> >> At this stage with -03 (which incorporates Ops and Sec reviews), we would >> like to request a review at Int-area. >> >> ICMP Extensions for Environmental Information >> draft-pignataro-eimpact-icmp-03 >> https://datatracker.ietf.org/doc/draft-pignataro-eimpact-icmp/ >> >> Thanks and regards, >> >> Carlos for the authors. >> _______________________________________________ >> Int-area mailing list -- [email protected] <mailto:[email protected]> >> To unsubscribe send an email to [email protected] >> <mailto:[email protected]>
signature.asc
Description: Message signed with OpenPGP
_______________________________________________ Int-area mailing list -- [email protected] To unsubscribe send an email to [email protected]
