[tcpm] Re: Some comments on rev -08 of draft-ietf-tcpm-tcp-ghost-acks

Michael Tuexen <michael.tuexen@lurchi.franken.de> Wed, 29 July 2026 17:28 UTC

Return-Path: <michael.tuexen@lurchi.franken.de>
X-Original-To: tcpm@mail2.ietf.org
Delivered-To: tcpm@mail2.ietf.org
Received: from localhost (localhost [127.0.0.1]) by mail2.ietf.org (Postfix) with ESMTP id 863BC120A5847; Wed, 29 Jul 2026 10:28:23 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=ietf.org; s=ietf1; t=1785346103; bh=7KtLmrtJCct6AGcEmvJZph4h8w2m6q5MB99qVgNtF1U=; h=Subject:From:In-Reply-To:Date:Cc:References:To; b=OoB2sieGI1GDC/aUMF5v5QmXACrKaVQMqF2zBBOx7Lb98VTMUrk0bq40YYYHsbN3E Us+CXu+amM+4SqPDj9b1YoeVxU6TspiaRj4Xo5Yegs1uQRBafhUPmuNr4x8dgg6fPV s+1xUjEZmlkIev/Ij+FRGL/u8qPd1KznSeiRX1Bc=
X-Virus-Scanned: amavisd-new at ietf.org
X-Spam-Flag: NO
X-Spam-Score: -1.097
X-Spam-Level:
X-Spam-Status: No, score=-1.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, RDNS_NONE=0.793, T_SPF_PERMERROR=0.01] autolearn=no autolearn_force=no
Received: from mail2.ietf.org ([166.84.6.31]) by localhost (mail2.ietf.org [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id OTN4vQWI1ym2; Wed, 29 Jul 2026 10:28:22 -0700 (PDT)
Received: from drew.franken.de (unknown [IPv6:2001:638:a02:a001:20e:cff:fe4a:feaa]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mail2.ietf.org (Postfix) with ESMTPS id 90402120A5844; Wed, 29 Jul 2026 10:28:22 -0700 (PDT)
Received: from smtpclient.apple (unknown [IPv6:2a02:8109:1101:be00:798a:1068:4511:94b5]) (Authenticated sender: lurchi) by drew.franken.de (Postfix) with ESMTPSA id 00B77721E2806; Wed, 29 Jul 2026 19:28:02 +0200 (CEST)
Content-Type: text/plain; charset="utf-8"
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3864.700.51.1.1\))
From: Michael Tuexen <michael.tuexen@lurchi.franken.de>
In-Reply-To: <a6d47266-f90b-452d-a085-3d14f24c3f5e@erg.abdn.ac.uk>
Date: Wed, 29 Jul 2026 19:28:02 +0200
Content-Transfer-Encoding: quoted-printable
Message-Id: <01848A87-B786-4CBC-A0F5-D6453C2A5711@lurchi.franken.de>
References: <3342128b-6c22-492e-8d5f-006999f5be59@erg.abdn.ac.uk> <47C61A2D-398D-44F6-88EC-F24F0006D095@lurchi.franken.de> <a6d47266-f90b-452d-a085-3d14f24c3f5e@erg.abdn.ac.uk>
To: Gorry Fairhurst <gorry@erg.abdn.ac.uk>
X-Mailer: Apple Mail (2.3864.700.51.1.1)
Message-ID-Hash: PHV7BTDMPC6PKXXIV43T2TCS4FG5SZ5I
X-Message-ID-Hash: PHV7BTDMPC6PKXXIV43T2TCS4FG5SZ5I
X-MailFrom: michael.tuexen@lurchi.franken.de
X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; emergency; loop; banned-address; member-moderation; header-match-tcpm.ietf.org-0; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header
CC: tcpm IETF list <tcpm@ietf.org>, "tcpm-chairs@ietf.org" <tcpm-chairs@ietf.org>
X-Mailman-Version: 3.3.9rc6
Precedence: list
Subject: [tcpm] Re: Some comments on rev -08 of draft-ietf-tcpm-tcp-ghost-acks
List-Id: TCP Maintenance and Minor Extensions Working Group <tcpm.ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/tcpm/nRpmSvnRUHLZqbnyb856tPzTPg0>
List-Archive: <https://mailarchive.ietf.org/arch/browse/tcpm>
List-Help: <mailto:tcpm-request@ietf.org?subject=help>
List-Owner: <mailto:tcpm-owner@ietf.org>
List-Post: <mailto:tcpm@ietf.org>
List-Subscribe: <mailto:tcpm-join@ietf.org>
List-Unsubscribe: <mailto:tcpm-leave@ietf.org>

> On 29. Jul 2026, at 15:06, Gorry Fairhurst <gorry@erg.abdn.ac.uk> wrote:
> 
> On 28/07/2026 16:13, Michael Tuexen wrote:]
> 
> See below, because I think something will need to change, but exactly which words depends on the way the document is finally presented.
>>> On 28. Jul 2026, at 14:44, Gorry Fairhurst <gorry@erg.abdn.ac.uk> wrote:
>>> 
>>> 
>>> I reviewed draft-ietf-tcpm-tcp-ghost-acks revision 08 as a part of an effort to provide input to the TCPM working group.
>>> 
>>> To me, teh current use of RFC 2119 language seems odd, because I read this as saying the specification will allow a conformant stack to use either approach described or no approach, and that seems like a poor choice for a standards-track document.  (e.g. “TCP stacks MAY implement one of the” - why “MAY”? )
>>> 
>>> Could the overall message be to say “ SHOULD implement the method in 3.2. When RFC4948 is not supported, a receiver can alternatively implement the method in 3.1?
>> What about using "TCP stacks SHOULD implement one of the" and leave the choice
>> it up to the implementer?
> 
> I don't think that is enoug to explain what its means to conform to this Spec.
> 
> i.e. Spec ought to require something, albeit it can an present two options about how to achieve it.
OK. What about:
"TCP stacks MUST implement one of the"
Would that resolve the issue?
> 
>> In my personal view section 3.1 is a complete description which only adds a boolean
>> variable to handle an arbitrary amount of bytes being sent. Section 3.2 describes
>> a method and leaves the handling of more than 2^64 byte up to the implementer.
>> 
>> I implemented the method in 3.1 in FreeBSD. As far as I know the method in 3.2 is
>> implemented in Linux, but without handling overflows.
> 
> - But is OK to do neither or something else (in which case what is still conformant?) So, I'd be happy with either method, to be sure, but allowing neither without clear requirements still seems at least very odd to me.
OK, I understand.
But wouldn't the above change not cover it?
> 
>>> I also wonder about the use of SHOULD in section 3.1 and 3.2... Then the following comments could simply be fixed by a minor rephrasing to use different RFC2119 language, but I’d like to be sure that I have correctly understood!
>> Thanks for bringing this up. Up to recently it was a mixture of MUST and SHOULD and non
>> normative language.
>>> * “At the beginning of the SEG.ACK validation, it SHOULD be determined”, I think this ought to be “MUST” when the receiver follows the method in 3.1? (If it really is a “SHOULD”, then what
>>> are the allowed conditions this approach is followed bye the check is not performed?)
>> The reason it is a SHOULD and not a MUST is that someone can implement
>> this pseudo code in a different way. So what about using a flag instead
>> of the boolean variable NO_ISS_CHECK.
> I am sure text can be found - and for sure the WG does not need to prescribe an implementation detail.
So what about removing the pseudo code and just state something like:

TCP stacks that implement this specification MUST add an input check that
the SEG.ACK value is acceptable only if a segment with such a SEG.SEQ has
been sent. All incoming segments whose SEG.ACK value doesn't satisfy this
condition MUST be discarded and a challege ACK sent back.
>>> * I think the same need for “MUST”  applies to the next two “SHOULD” clauses in section 3.1 - I wonder could all of these be “MUST” when a receifollowing 3.1?
>> What is about arranging the statements in a different way which is
>> performing better, but logically equivalent?
>> 
>> I think it is important to be consistent in Section 3.1 and Section 3.2.
>> The reason why SHOULD is used instead of MUST is given above.
>> But if you
>> think the authors need to replace SHOULD with MUST I am fine with it,
> See above :-)
>>  as
>> long as the pseudo code is not binding.
> Exactly, that really ought not to be binding.
So maybe not providing any code at all.
>>> *  I also then wonder if “SHOULD ”,  ought to be “MUST” when following the approach in 3.2?
>> This should be consistent with section 3.1.
>>> * In 3.2  “An implementation SHOULD handle tcpEStatsAppHCThruOctetsAcked overflow.” - is this a requirement, or is there a reason that complaint implementation will not check this — if so this needs to be explained, and if it remains a “SHOULD” is this a concern when not implemented?
>> Well, as far as I know (see Acknowledgment section) this method is what is used in
>> Linux. Originally, there was no handling of overflows discussed. This came in during
>> discussion in the WG with a MAY and was finally changed to a SHOULD. I don't know if Linux protects the counter against overflowing or not. The actual check is not protected.
> 
> In summary, my take would be:
> 
> - If you conform to this RFC-to-be, then you MUST do "something" and say what.
I agree.
> 
> - I super-agree that implementation details ought not to be a requirement and there can be a choice: I have no insight to say that one method is needed rather than other, that seems like an implementation choice between two options?
Yes.
> 
> - I think the consequences of following all the SHOULDs needs to be explained
So what about avoiding pseudo code at all? Would you prefer that?

Best regards
Michael
> 
> Gorry
> 
>> 
>> Best regards
>> Michael (as an individual)
>>> Best wishes,
>>> Gorry
>>> (as a member of TCPM WG)
>>> 
>>> 
>>> P.S. I have a few editorial comments:
>>> 
>>> 1. is it true that “TCP as specified in [RFC0793] is widely deployed in today's Internet.”? Maybe an explanation or reference is needed to support this claim? (Although, I’m unsure if this claiming support for sept in RFC793, or just TCP support - is this perhaps what was intended: “TCP (as originally specified by [RFC0793] is widely supported…”
>>> 2. “This document aims to improve” … This seems like there is doubt, may be it could be “This document improves”.
>>> 3. “has never sent.” - is there an extra paragraph break or (I think) a missing space before the next paragraph?
>>> 
>>> _______________________________________________
>>> tcpm mailing list -- tcpm@ietf.org
>>> To unsubscribe send an email to tcpm-leave@ietf.org
> 
>