Top post, historic view. IIRC the reason for not requiring a Notification in the case of overload was that the process of sending a Notification might contribute to the overload. And, furthermore, that there might be an attack that leverages the need to send a Notification to perpetuate the overload.
I take no position on this: just reporting what is in my memory. Best, Adrian -----Original Message----- From: Dhruv Dhody <[email protected]> Sent: 16 September 2019 12:21 To: Barry Leiba <[email protected]>; [email protected] Cc: The IESG <[email protected]>; pce-chairs <[email protected]>; Farrel Adrian <[email protected]>; [email protected] Subject: Re: **Barry Leiba's DISCUSS on draft-ietf-pce-stateful-pce-auto-bandwidth** Hi again! On Mon, Sep 16, 2019 at 4:48 PM Dhruv Dhody <[email protected]> wrote: > > Hi Barry, WG, > > I saw the DISCUSS [1] in the datatracker but for some reason the email > never landed in my inbox or the list [2]. I am manually posting it > here - > > ==== > > Discuss (2019-09-16) > > Thanks for another clear document. There are some "SHOULD" key words > in one section that I would like to discuss, and that I think we ought > to be able to resolve without much difficulty: > > — Section 5.7 — > > There are various “SHOULD”s in this section, and I have the same > comment about all of them: BCP 14 says, about “SHOULD”, that “there > may exist valid reasons in particular circumstances to ignore a > particular item, but the full implications must be understood and > carefully weighed before choosing a different course.” I see no > guidance here to help the reader understand what such circumstances > and implications are, so I can’t see how an implementer can evaluate > the situation. Can you provide any help here? > > ==== > I checked the base RFC for PCEP - RFC 5440 where notifications are first defined. They do not use MUST for sending notification in the PCE overload case [1]. Leaving that aside, in case of auto-bandwidth feature, this notification is important for scaling. I am inclined to change it to MUST as suggested. Co-authors, WG, please speak up if you disagree!! I have incorporated all other comments in the working copy. Diff: https://tools.ietf.org/tools/rfcdiff/rfcdiff.pyht?url1=draft-ietf-pce-stateful-pce-auto-bandwidth-11&url2=https://raw.githubusercontent.com/dhruvdhody/ietf/master/draft-ietf-pce-stateful-pce-auto-bandwidth-12.txt Thanks! Dhruv [1] https://tools.ietf.org/html/rfc5440#section-7.14 > Comment (2019-09-16) > > Again, these are purely editorial comments, which need no detailed > response; please just consider them. > > — Section 1 — > > Over time, based on the varying traffic pattern, an LSP established > with a certain bandwidth may require to adjust the bandwidth reserved > in the network dynamically. > > “may require adjustment of the bandwidth” > > This is similar to > the Passive stateful PCE model, while the Passive stateful PCE uses > path request/reply mechanism, the Active stateful PCE uses > report/update mechanism. > > NEW > This is similar to > the Passive stateful PCE model: while the Passive stateful PCE uses > a path request/reply mechanism, the Active stateful PCE uses a > report/update mechanism. > END > > This document defines the PCEP extensions needed to support Auto- > Bandwidth feature in a Active stateful PCE model > > NEW > This document defines the PCEP extensions needed to support an Auto- > Bandwidth feature in an Active stateful PCE model > END > > — Section 2.3 — > > This value indicates how many times > consecutively, the percentage or absolute difference > > Add a comma after “times”. > > — Section 3 — > > The PCEP speaker supporting this document must have a mechanism > > “A PCEP speaker”. > > o It is required to identify and inform the PCC, which LSPs are > enabled with Auto-Bandwidth feature. Not all LSPs in some > deployments would like their bandwidth to be dependent on the > real-time bandwidth usage but be constant as set by the operator. > > NEW > o It is necessary to identify and inform the PCC which LSPs have > the Auto-Bandwidth feature enabled. In some deployments, not > all LSPs would like their bandwidth to be dependent on the > real-time bandwidth usage, but would rather be constant as set > by the operator. > END > > — Section 4.1 — > > The initial LSP bandwidth can be set to an arbitrary value (including > zero), in practice, it can be operator expected value based on design > and planning. > > NEW > The initial LSP bandwidth can be set to an arbitrary value (including > zero). In practice, it can be set to an expected value based on design > and planning. > END > > — Section 4.2 — > > When the Auto-Bandwidth feature is enabled, the measured traffic rate > is periodically sampled at each Sample-Interval (which can be > configured by an operator and the default value as 5 minutes) by the > PCC, when the PCC is the head-end node of the LSP. The traffic rate > samples are accumulated over the Adjustment-Interval period (in the > Up or Down direction) (which can be configured by an operator and the > default value as 24 hours). > > NEW > When the Auto-Bandwidth feature is enabled, the measured traffic rate > is periodically sampled at each Sample-Interval by the PCC, when the > PCC is the head-end node of the LSP. The sample interval can be > configured by an operator, with a default value of 5 minutes. > > The traffic rate samples are accumulated over the Adjustment-Interval > period (in the Up or Down direction). The period can be configured by > an operator, with a default value of 24 hours. > END > > The PCC, in-charge of calculating the > bandwidth to be adjusted, can decide to adjust the bandwidth > > Remove both commas. > > Only if the difference between the > current bandwidth demand (MaxAvgBw) and the current bandwidth > reservation is greater than or equal to the Adjustment-Threshold > (percentage or absolute value) (which can be configured by an > operator and the default as 5 percentage), the LSP bandwidth is > adjusted (upsized) to the current bandwidth demand (MaxAvgBw). > > I’m sorry: I can’t made any sense out of this text and, thus, can’t > suggest a fix. Please try rephrasing this. When you do, please make > it more than one sentence, and please avoid consecutive parenthesized > phrases, which are awkward. > > However, longer > adjustment-interval can result in an undesirable effect > > “a longer” > > To avoid this, the > Auto-Bandwidth feature may pre-maturely expire the adjustment- > interval and adjust the LSP bandwidth > > “prematurely”, with no hyphen. > “adjustment interval”, with no hyphen. > > — Section 5.1 — > > o The PCEP speaker that does not recognize the extensions defined in > > “A PCEP speaker” > > o If the PCEP speaker that supports the extensions defined in this > > “If a PCEP speaker supports” > > — Section 5.2 — > > Future specification can define additional sub-TLVs. > > “specifications” > > If sub-TLVs are not present, the > default values as specified in this document are used or otherwise > based on the local policy are assumed. > > I can’t make sense of that sentence; please re-phrase it. > > — Section 5.2.3.2 — > > o Reserved: SHOULD be set to zero on transmission and MUST be > ignored on receipt. > > Why is this “SHOULD”, when other reserved values have been “MUST”? > > (Same comment in 5.2.3.4, 5.2.5.1, 5.2.5.2, 5.2.5.3, and 5.2.5.4.) > > ==== > > [1] > https://datatracker.ietf.org/doc/draft-ietf-pce-stateful-pce-auto-bandwidth/ballot/ > [2] https://mailarchive.ietf.org/arch/browse/pce/ _______________________________________________ Pce mailing list [email protected] https://www.ietf.org/mailman/listinfo/pce
