Document: draft-ietf-bmwg-mlrsearch Title: Multiple Loss Ratio Search Reviewer: Jen Linkova Review result: Ready with Nits I have reviewed this document as part of the INT directorate's ongoing effort to review all IETF documents being processed by the IESG. These comments were written primarily for the benefit of the INT area directors. Document editors and WG chairs should treat these comments just like any other last call comments. This informational document defines a new methodology called Multiple Loss Ratio search (MLRsearch) for throughput benchmarking. I think the document does not introduce any new concerns, especially ones which might be of interest for the INT ADs. However, as I have very limited knowledge of the topic, I have some rather optional comments from an uneducated reader point of view. I hope at least some of those comments make sense and might improve the document readability. My main overall comment is that terminology might present some challenges for an uneducated reader: some terms are used before they defined, and it's not always easy to find the exact definitions. A few examples: 1. Introduction says: "Applying the vanilla throughput binary search,as specified for example in [TST009] and [RFC2544]" RFC2544 mention 'search' exactly twice in Section 24 but doesn't really says much. IMHO it might be benefitial to refer to Section 12.3.2 of TST009 and/or section 5.1 of this document. 2. Introduction (and other sections) uses the term "Search Goal" extensively, like it's a well-known term, but then Section 4.6.7 explains that it is actually a new term, introduced by this document. IMHO, adding a reference to Section 4.6.7 when that term used for the very first would be very useful. Also, I feel like Section 1.2 kind of overlaps with 4.1 Taking this into account, I'm wondering if the following changes might improve the document readability: 1. Moving the content of Section 1.2 to Section 4.1 (if it's not already there). 2. Add a Terminology section right after the introduction, and say smth similar to what the very beginnging of Section 4.4 is saying: "please read the following documents, as this document reuses the terminology" and also mention that a lot of terms are defined in Sections 4.3 - 4.9? (Requirement Langucage could be the next section, right after it). Some other nits: Introduction: "Be careful when dealing with inconsistent trial results" - how about "Process of inconsistent trial results more carefully" - so it matches how other bullet points are phrased? Also, it's a bit unclear what the first subpoint is saying. "Reported throughput is smaller than the smallest load with high loss." - is it "to ensure that repored throughput is smaller..."? (I'm not a native speaker but IMHO "lower throughput" sounds a bit better). "MLRsearch configuration supports both conservative settings and aggressive settings. Conservative enough settings lead to results unconditionally compliant with [RFC2544]," "Conservaive enough" suggests there is a spectrum of settings. However, if I unserstand Section 4.10.2 correctly, there are specific well defined values which ensure that the rpopsed algorith behaves as per RFC2544. Maybe it's worth rephrasing, smth like: "MLRsearch configuration supports variety of settings. The more conservative the settings are, the close the MLRsearch behavioir to RFC2544 (see Section 4.10.2 for more details)." Section 2.2 Would be nice to expland SUT acronym upon its first use in the text, not later. Section 2.4 The last bullet point in the "Motivations are many:" list shouldn't actually be a bullet point, as it's not "one of many". "Regardless of the validity of all similar motivations, support for non-zero loss goals makes a search algorithm more user-friendly. [RFC2544] throughput is not user-friendly in this regard." IMHO this needs clarification, as it's a bit unclear what "user-friendly" means in this context, especially - as the first sentence seems to imply - it's unrelated to all the motivations listed above? "At the time of writing there does not seem to be a consensus in the industry on which ratio value is the best." I assume "ratio" means "loss ratio", worth making it explicit. Also, IMHO, it should be noted that the user experience would vary - even for a given loss ration - depending on the specific upper-layer protocol and the application. Section 4.2 Another case of terms used before the definition. When you introduce the terms "regular" and "irregular" seach results, it owuld be benefitial to refer to section 4.8.4, where they are defined. Section 5.3.2 "Depending on Search Goal attributes, Load Classification results may be resistant to small amounts of Section Inconsistent Trial Results (Section 2.5)." - I assume it's a typo, and should be smth like "...to small amount of inconsistencies between trial results, Section 2.5)" There are a number of places where a new paragraph starts in the middle of thes entence, in particular: Sections 2.1, 5.3, 5.3.2, 5.4 -- last-call mailing list -- last-call@xxxxxxxx To unsubscribe send an email to last-call-leave@xxxxxxxx
