Hi Gorry and Neal,
Thank you for your replies. We have updated the files accordingly.
2. Section 4, CURRENT:
However, note that using PRR for cwnd reductions
for the [RFC3168] variant of ECN has been observed, with some
approaches to Active Queue Management (AQM), to cause an excess cwnd
reduction during ECN-triggered congestion episodes, as noted in
[VCC].
NEW:
However, there can be interactions between using PRR and
approaches to Active Queue Management (AQM) and ECN; guidance on the
development and assessment of congestion control
mechanisms is provided in [RFC9743].
- RFC-Ed: Pleaase remove the reference to [VCC] from the references
section.
) FYI, we have added [RFC9743] as an Informative Reference. Please let us
know if it should be a Normative Reference instead.
The files have been posted here (please refresh):
https://www.rfc-editor.org/authors/rfc9937.txt
https://www.rfc-editor.org/authors/rfc9937.pdf
https://www.rfc-editor.org/authors/rfc9937.html
https://www.rfc-editor.org/authors/rfc9937.xml
The relevant diff files are posted here:
https://www.rfc-editor.org/authors/rfc9937-diff.html (comprehensive diff)
https://www.rfc-editor.org/authors/rfc9937-auth48diff.html (all AUTH48
changes)
https://www.rfc-editor.org/authors/rfc9937-lastdiff.html (last version to
this one)
https://www.rfc-editor.org/authors/rfc9937-lastrfcdiff.html (rfcdiff
between last version and this)
Please see the AUTH48 status page for this document here:
https://www.rfc-editor.org/auth48/rfc9937
We will await approval from Nandita prior to moving this document forward
in the publication process.
Best regards,
Alanna Paloma
RFC Production Center
On Dec 15, 2025, at 9:50 AM, Gorry Fairhurst<[email protected]>
wrote:
RFC-Ed, as per my previous email, here is the consolidated change list,
please go-ahead and apply these changes:
1. Section 4, CURRENT:
such as for Explicit Congestion Notification (ECN) as
described in [RFC3168].
NEW:
such as for Explicit Congestion Notification (ECN) as
specified in [RFC3168].
2. Section 4, CURRENT:
However, note that using PRR for cwnd reductions
for the [RFC3168] variant of ECN has been observed, with some
approaches to Active Queue Management (AQM), to cause an excess cwnd
reduction during ECN-triggered congestion episodes, as noted in
[VCC].
NEW:
However, there can be interactions between using PRR and
approaches to Active Queue Management (AQM) and ECN; guidance on the
development and assessment of congestion control
mechanisms is provided in [RFC9743].
- RFC-Ed: Pleaase remove the reference to [VCC] from the references
section.
3. Section 11.2, CURRENT:
PRR only operates
during a congestion control response episode, such as fast recovery
or response to the [RFC3168] variant of ECN,
NEW:
PRR only operates
during a congestion control response episode, such as fast recovery
or when there is a step reduction in the cwnd from the TCP ECN
reaction defined in [RFC3168].
-END-
Gorry
(WIT AD)
On Dec 11, 2025, at 9:09 AM, Yuchung Cheng<[email protected]>
wrote:
Looks good to me. I approve the latest draft.
On Wed, Dec 10, 2025 at 7:59 PM Neal Cardwell<[email protected]>
wrote:
Hi Alanna,
Thanks! Looks great to me.
I approve of this draft posted today, December 10.
Everybody else, please review and chime in.
Thanks!
neal
On Wed, Dec 10, 2025 at 8:01 PM Alanna Paloma <
[email protected]> wrote:
Hi Neal,
The files have been updated per your request.
The files have been posted here (please refresh):
https://www.rfc-editor.org/authors/rfc9937.txt
https://www.rfc-editor.org/authors/rfc9937.pdf
https://www.rfc-editor.org/authors/rfc9937.html
https://www.rfc-editor.org/authors/rfc9937.xml
The relevant diff files are posted here:
https://www.rfc-editor.org/authors/rfc9937-diff.html (comprehensive
diff)
https://www.rfc-editor.org/authors/rfc9937-auth48diff.html (all
AUTH48 changes)
https://www.rfc-editor.org/authors/rfc9937-lastdiff.html (last
version to this one)
https://www.rfc-editor.org/authors/rfc9937-lastrfcdiff.html
(rfcdiff between last version and this)
Please see the AUTH48 status page for this document here:
https://www.rfc-editor.org/auth48/rfc9937
We will await approvals from each party listed on the AUTH48 status
page below prior to moving this document forward in the publication process.
Thank you,
Alanna Paloma
RFC Production Center
On Dec 10, 2025, at 3:59 PM, Neal Cardwell<[email protected]>
wrote:
Hi Alanna,
I had 4 minor editing requests based on the December 3 version of
the text:
--- in "4. Changes Relative to RFC 6937":
rfc6937bis-21:
using PRR for cwnd reductions for [RFC3168] ECN
OLD:
using PRR for cwnd reductions for ECN [RFC3168]
NEW:
using PRR for cwnd reductions for the [RFC3168] variant of ECN
Rationale: IMHO the auth48 edit to use the phrase "ECN [RFC3168]"
implies that there is only one version of ECN. However, there are at least
3: classic [RFC3168], DCTCP [RFC8257], and L4S [RFC9331]. Here [RFC3168]
is intended as an adjective clarifying which flavor of ECN we are
discussing, not to indicate that ECN is only defined in [RFC3168].
I'd suggest using the "the [RFC3168] variant of ECN" phrase that is
currently in Section "11.2. Fairness".
---- in 6.2. Per-ACK Steps
rfc6937bis-21:
Finally, the sender uses DeliveredData, inflight, SafeACK, and
other
PRR state to compute
SndCnt, a local variable indicating exactly how
many bytes should be sent in response to each ACK,
and then uses SndCnt to update cwnd
OLD:
Finally, the sender uses DeliveredData, inflight, SafeACK, and
other
PRR state to compute
SndCnt, a local variable indicating exactly how
many bytes should be sent in response to each ACK
and then uses SndCnt to update cwnd
NEW:
Finally, the sender uses DeliveredData, inflight, SafeACK, and
other
PRR state to compute
SndCnt, a local variable indicating exactly how
many bytes should be sent in response to each ACK,
and then uses SndCnt to update cwnd
Rationale: the phrase "a local variable indicating exactly how many
bytes should be sent in response to each ACK" is a parenthetic or
non-restrictive clause, so AFAIK should be enclosed with commas before and
after. (Strunk & White Elements of Style rule: "Enclose parenthetic
expressions between commas".)
--- in 8. Examples
rfc6937bis-21:
This section illustrates the PRR and [RFC6675] algorithms
OLD:
This section illustrates the PRR and [RFC6675] algorithm
NEW:
This section illustrates the PRR and [RFC6675] algorithms
Rationale:
PRR and [RFC6675] are two different algorithms.
--- in 14.2. Informative References
rfc6937bis-21:
[FACK] Mathis, M. and J. Mahdavi, "Forward Acknowledgment:
Refining TCP Congestion Control", ACM SIGCOMM
SIGCOMM1996, August 1996,
OLD:
[FACK] Mathis, M. and J. Mahdavi, "Forward Acknowledgment:
Refining TCP Congestion Control", ACM SIGCOMM
Computer
Communication Review, vol. 26, no. 4, pp. 281-291,
NEW:
[FACK] Mathis, M. and J. Mahdavi, "Forward Acknowledgment:
Refining TCP Congestion Control", SIGCOMM '96:
Conference
Proceedings on Applications, Technologies,
Architectures,
and Protocols for Computer Communications, pp.
281-291,
Rationale: IMHO it's very useful/important to indicate that a paper
is a SIGCOMM paper, so we should not drop the fact that the FACK paper was
in SIGCOMM '96 (the list of SIGCOMM '96 papers is here:
https://dl.acm.org/doi/proceedings/10.1145/248156 ). I'm suggesting the
"NEW" text for indicating the paper was in SIGCOMM '96 based on the fact
that [Hoe96Startup] was also in SIGCOMM '96; so I've just borrowed the
SIGCOMM '96 citation text from [Hoe96Startup], which looks like:
[Hoe96Startup]
Hoe, J., "Improving the Start-up Behavior of a Congestion
Control Scheme for TCP", SIGCOMM '96: Conference
Proceedings on Applications, Technologies, Architectures,
and Protocols for Computer Communications, pp. 270-280,
DOI 10.1145/248157.248180, August 1996,
<https://doi.org/10.1145/248157.248180>.
(The confusion arises because SIGCOMM papers can be cited in two
ways: (1) as in the SIGCOMM 'XY conference proceedings, or (2) as an issue
of ACM SIGCOMM Computer Communication Review.)
Thanks!
neal
On Wed, Dec 10, 2025 at 5:14 PM Alanna Paloma <
[email protected]> wrote:
Hi Matt,
Thank you for your approval. It’s been noted on the AUTH48 status
page:
https://www.rfc-editor.org/auth48/rfc9937
We will await approvals from Neal, Yuchung, and Nandita prior to
moving this document forward in the publication process.
Best regards,
Alanna Paloma
RFC Production Center
On Dec 10, 2025, at 7:54 AM, Matt Mathis<[email protected]>
wrote:
I approve of the draft posed Dec 3rd.
Everybody else, please review and chime in.
Thanks,
--MM--
Evil is defined by mortals who think they know "The Truth" and use
force to apply it to others.
-------------------------------------------
Matt Mathis (Email is best)
Home & mobile: 412-654-7529 <(412)%20654-7529> please leave a
message if you must call.
On Wed, Dec 3, 2025 at 12:26 PM Alanna Paloma <
[email protected]> wrote:
Hi Neal and Gorry,
Thank you for your replies. Gorry’s approval has been noted on the
AUTH48 status page:
https://www.rfc-editor.org/auth48/rfc9937
One preliminary meta-note about process:
https://www.rfc-editor.org/authors/rfc9937-auth48diff.html
(all AUTH48 changes)
FWIW, AFAICT this version does not include all auth48 changes.
One change I noticed that it does not include is the following:
rfc6937bis-21:
using [RFC6675] loss detection
MAY use the "pipe" algorithm as specified in [RFC6675]
latest auth48 version:
using loss detection [RFC6675]
MAY use the "pipe" algorithm as specified in [RFC6675]
) To clarify, the -auth48diff file only highlights changes after a
document has moved into AUTH48 state. The change you noted was not
highlighted in the -auth48diff file (it's now highlighted as we have
reverted our initial edit per your request) because it was made by editors
prior to the document entering AUTH48 state.
To see all edits made, including those made before and during
AUTH48 state, see this file:
https://www.rfc-editor.org/authors/rfc9937-diff.html
(comprehensive diff)
We have updated the files per your request.
The files have been posted here (please refresh):
https://www.rfc-editor.org/authors/rfc9937.txt
https://www.rfc-editor.org/authors/rfc9937.pdf
https://www.rfc-editor.org/authors/rfc9937.html
https://www.rfc-editor.org/authors/rfc9937.xml
The relevant diff files are posted here:
https://www.rfc-editor.org/authors/rfc9937-diff.html
(comprehensive diff)
https://www.rfc-editor.org/authors/rfc9937-auth48diff.html (all
AUTH48 changes)
https://www.rfc-editor.org/authors/rfc9937-lastdiff.html (last
version to this one)
https://www.rfc-editor.org/authors/rfc9937-lastrfcdiff.html
(rfcdiff between last version and this)
We will await any further changes you may have and approvals from
each author prior to moving forward in the publication process.
Please see the AUTH48 status page for this document here:
https://www.rfc-editor.org/auth48/rfc9937
Thank you,
Alanna Paloma
RFC Production Center
On Dec 3, 2025, at 6:29 AM, Neal Cardwell<[email protected]>
wrote:
Hi editors and co-authors,
I had time to review the auth48 edits this morning, and have some
proposed edits.
One preliminary meta-note about process:
https://www.rfc-editor.org/authors/rfc9937-auth48diff.html
(all AUTH48 changes)
FWIW, AFAICT this version does not include all auth48 changes.
One change I noticed that it does not include is the following:
rfc6937bis-21:
using [RFC6675] loss detection
MAY use the "pipe" algorithm as specified in [RFC6675]
latest auth48 version:
using loss detection [RFC6675]
MAY use the "pipe" algorithm as specified in [RFC6675]
Here are a few edits I'd like to request, tweaking the edits made
during the auth48 process:
---
rfc6937bis-21:
using [RFC6675] loss detection
MAY use the "pipe" algorithm as specified in [RFC6675]
OLD:
using loss detection [RFC6675]
MAY use the "pipe" algorithm as specified in [RFC6675]
NEW:
using [RFC6675] loss detection
MAY use the "pipe" algorithm as specified in [RFC6675]
Rationale: IMHO the auth48 edit to employ the phrase "using loss
detection [RFC6675]" implies that loss detection necessarily means
[RFC6675], or is only defined in [RFC6675]. However, there are multiple
widely-deployed loss recovery algorithms (notably [RFC6675] and [RFC8985]),
and this paragraph we are specifically discussing how to adapt PRR's use
of the "inflight" quantity to both of those algorithms, and in this
sentence we are discussing how to adapt PRR's use of the "inflight"
quantity to [RFC6675] loss detection, so it's important not to imply that
loss detection is only defined in [RFC6675].
---
rfc6937bis-21:
Finally, the sender uses DeliveredData, inflight, SafeACK, and
other
PRR state to compute SndCnt
OLD:
Finally, the sender uses DeliveredData, inflight, SafeACK, and
other
PRR states to compute SndCnt
NEW:
Finally, the sender uses DeliveredData, inflight, SafeACK, and
other
PRR state to compute SndCnt
Rationale: IMHO the auth48 edit to use "states" implies that
DeliveredData, inflight, SafeACK are names of "states" in a state machine
