[tcpm] Re: New Version Notification for draft-bonica-tcpm-extended-options-04.txt
Michael Tuexen <michael.tuexen@lurchi.franken.de> Tue, 04 August 2026 14:56 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 267991236ED80 for <tcpm@mail2.ietf.org>; Tue, 4 Aug 2026 07:56:50 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=ietf.org; s=ietf1; t=1785855410; bh=HNyHXOn9IP9DFwkxSznv6AwfwO5l8ymTh+tuRVb2Tlo=; h=Subject:From:In-Reply-To:Date:Cc:References:To; b=yCDKvKUkj92TAnwawpO7mJMWPfzG9IwSwLnBpJvadxagOie8Uy2UVjJsm8+DRiEVb 5DS62btNBX3LgbfqtLXvGgNnVdLHktT1vtmgf45FGC7AWssRgD7CdhCiLrKSqBpVRg gLCe1PbhlRwaph+V9JBh2YuLp+g20eWMSjP3Bi5c=
X-Virus-Scanned: amavisd-new at ietf.org
X-Spam-Flag: NO
X-Spam-Score: -2.589
X-Spam-Level:
X-Spam-Status: No, score=-2.589 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, RCVD_IN_DNSWL_LOW=-0.7, RCVD_IN_VALIDITY_CERTIFIED_BLOCKED=0.001, RCVD_IN_VALIDITY_RPBL_BLOCKED=0.001, SPF_PASS=-0.001, T_PDS_SHORTFWD_URISHRT_QP=0.01] autolearn=ham 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 LGWOqFSZgYep for <tcpm@mail2.ietf.org>; Tue, 4 Aug 2026 07:56:49 -0700 (PDT)
Received: from drew.franken.de (mail-n.franken.de [193.175.24.27]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mail2.ietf.org (Postfix) with ESMTPS id E6E6C1236ED78 for <tcpm@ietf.org>; Tue, 4 Aug 2026 07:56:48 -0700 (PDT)
Received: from smtpclient.apple (unknown [IPv6:2a02:8109:1101:be00:f0cd:3fc0:3cb2:76e5]) (Authenticated sender: lurchi) by drew.franken.de (Postfix) with ESMTPSA id C783C721E283A; Tue, 4 Aug 2026 16:56:35 +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: <60466753-B2AB-4FB8-81C8-2DA6C021F225@beachpine.com>
Date: Tue, 04 Aug 2026 16:56:35 +0200
Content-Transfer-Encoding: quoted-printable
Message-Id: <D04EF330-5A58-47C0-B39A-F86E3ED3E52E@lurchi.franken.de>
References: <178247435258.1809941.9081592875087275710@dt-datatracker-f9b87776f-8pmmg> <DM4PR84MB231000F604E508DE0EAEF63FF4EB2@DM4PR84MB2310.NAMPRD84.PROD.OUTLOOK.COM> <3c6ef75b-0594-46b4-b748-478b48013a54@gmx.at> <DM6PR09MB5958A0C4E2F68D2ED19631749FD72@DM6PR09MB5958.namprd09.prod.outlook.com> <CACL_3VH9T_mYREMF0JS6i_NqML-9K-TMCh-S57K2xzENDh+TXQ@mail.gmail.com> <FB9E95B7-CE73-4A63-9121-AF49D582F78B@beachpine.com> <CACL_3VGJvVsjk3yBcECdoXFHZuMomTxYg05Chfd4VdQgU9fdOA@mail.gmail.com> <60466753-B2AB-4FB8-81C8-2DA6C021F225@beachpine.com>
To: Joe Touch <touch@beachpine.com>
X-Mailer: Apple Mail (2.3864.700.51.1.1)
Message-ID-Hash: NVYLAUB3ZQNVSGNFQHK57RGHIVFGRB2C
X-Message-ID-Hash: NVYLAUB3ZQNVSGNFQHK57RGHIVFGRB2C
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 <tcpm@ietf.org>, Tony Li <tony1athome@gmail.com>, "Bonica, Ron" <ronald.bonica@hpe.com>
X-Mailman-Version: 3.3.9rc6
Precedence: list
Subject: [tcpm] Re: New Version Notification for draft-bonica-tcpm-extended-options-04.txt
List-Id: TCP Maintenance and Minor Extensions Working Group <tcpm.ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/tcpm/LfVYhLkOhUunVoKr8smft8NuTPg>
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 4. Aug 2026, at 02:51, Joe Touch <touch@beachpine.com> wrote: > > Some notes below... > >> On Aug 2, 2026, at 4:40 PM, C. M. Heard <heard@pobox.com> wrote: >> >> Greetings everyone, >> >> Please see responses inline below. >> >> On Fri, Jul 31, 2026 at 9:08 PM Joe Touch <touch@beachpine.com> wrote: >> It’s difficult to understand the logic below, given tcp-syn-ext-opt uses a packet format that also is silently ignored by legacy endpoints. They’re simply ignored in different ways. >> >> Syn-ext-opt uses a packet that is not defined by TCP, thus is implicitly not invalid, but not otherwise defined. >> >> That's true, and I admit that I had not noticed that. However, in my defense, I intended the main thrust of my comments to contrast differences between the approaches in draft-ietf-tcpm-tcp-edo and draft-bonica-tcpm-extended-options. >> Ron’s method defines a packet format that is invalid by current TCP standards, thus - IMO - much more likely to trigger unexpected behavior, e.g., from accelerators and other devices. >> >> Actually, I think that the opposite is likely to be true. If those accelerators and other devices don't understand the EDO option but still attempt to participate in the connection, and they assume that RFC 9293 means what it says about unknown options, i.e., requirement MUST-6: >> >> Feature >> ReqID >> MUST >> SHOULD >> MAY >> SHOULD NOT >> MUST NOT >> TCP Options >> Support the mandatory option set >> MUST-4 >> X >> Receive TCP Option in any segment >> MUST-5 >> X >> Ignore unsupported options >> MUST-6 >> X >> >> then they will misinterpret any TCP packets with an EDO option specifying a data offset larger than that specified in the main TCP header, resulting in corrupted data being passed to the user. That's definitely a bad outcome. > > > Agreed. And that should have been caught in that RFC; this “ignore unsupported options” is something in direct violation of TCP’s core spec. I don't understand this point. RFC 9293 contains: A TCP implementation MUST (MUST-6) ignore without error any TCP Option it does not implement, assuming that the option has a length field. Doesn't this contradict your statement? > >> On the other hand, although this is admittedly not an explicit requirement of RFC 9293, it is commonly assumed that the great majority of implementations will check the data offset (DO) field in the main TCP header for basic format errors and thus will, as a rule, silently discard packets with DO=0. In this case the connection fails entirely, but at least passing corrupted data to the user does not happen. > > > That’s a bug not a feature. It also means that legacy endpoints won’t respond - won’t send a RST, won’t anything. They just hang as non responsive. > > That begs the question whether this is actually a TCP option or really, as I originally contended, something that grifts off TCP’s codebase but isn’t TCP. > > I would say that if you’re sending packets that legacy TCP ignores, you’re not TCP - why not just run QUIC at that point? > >> I do understand that this benign behavior cannot be assumed to be universal: during the recent IETF meeting, Richard Scheffenegger pointed out at the mic that certain implementations are known to crash when receiving packets with DO=0 (see https://youtu.be/pS7dKMeaHeE?t=2064) and as he says, if an experiment based on draft-bonica-tcpm-extended-options is conducted, it's probably wise to determine how prevalent this is. And though this is not really an issue for draft-bonica-tcpm-extended-options, it may also be incumbent on the WG to publish an update to RFC 9293 with text modelled after that in RFC 1814 Section 5.2.2 (I admit that I was surprised not to find this in the RFC already). To be perfectly frank, that crashing behavior reminds me somewhat of the Ping O'Death bug that was found around 30 years ago to be widespread, but once found it was quickly fixed in mainstream implementations. Barring extraordinary circumstances, I'm not very sympathetic to suggestions that useful innovations should be stifled by the existence of defective TCP implementations that don't check the DO field. >> >> After the initial handshake, they both use similar “ignore the header length, use this instead” option in subsequent packets and both have the same underlying issue that prevented EDO from moving forward - they are always still incompatible with TCP accelerators, which violate TCP semantics by aggregating packets even for options they do not understand. >> >> In fact, draft-bonica-tcpm-extended-options specifically required the use of the DO=0 format during the three-way handshake, and that is where its big operational advantage over EDO comes from: it can accommodate long options in every packet, including the initial SYN, SYN-ACK, and ACK. > > > And that is also its bug; it won’t cannot back off to legacy TCP. EDO’s SYN option can - if the surplus packet is late or never arrives. > > FWIW, this also means that this “not-TCP-TCP” won’t do NAT or even open a triggered-firewall properly. Legacy implementations would just black-hole the packets. > > We’ve been around the black-hole problem for a while; creating a new way to do that seems counterproductive. > >> And no, EDO has not been abandoned per se - it’s just silly to waste time doing in-place updates of a draft when the draft hasn’t changed. >> >> Understood, and I was not claiming that. Rather, I am suggesting that it has been overtaken by events. > > > I disagree. EDO remains the only legacy-compatible way to extend options, and it depends on what events you refer to. The arrival of this non-TCP-TCP doesn’t magically make EDO not legacy compatible; it raises more issues than it solves. > > Finally, there is the issue about what happens when either of these “fail”, i.e., when something incorrectly ignores them. Magic numbers won’t save anyone; the appropriate approach is to sign the packet if integrity needs to be verified. > > Joe > >> Thanks, >> >> Mike Heard >> >> >> Joe >> >>> On Jul 31, 2026, at 8:52 PM, C. M. Heard <heard@pobox.com> wrote: >>> >>> Greetings TCPM, >>> >>> I would like to advocate for a view somewhat at odds with the one that RichardScheffenegger presents below. Specifically, I see EDO and similar approaches that strive to maintain backward compatibility with legacy implementations as a dead end that ought to be abandoned after decades of failure to gain traction. Rather than attempting to do both EDO (possibly augmented by something like draft-touch-tcpm-tcp-syn-ext-opt) and the approach in draft-bonica-tcpm-extended-options, the WG should pick one approach and run with it. I am admittedly an outsider to the TCP world, but I see the approach in draft-bonica-tcpm-extended-options as technically superior to that of EDO and its companion draft-touch-tcpm-tcp-syn-ext-opt. Slide #7 of Ron's presentation sums it up pretty well: >>> >>> •TCP-O and TCP-U are incompatible >>> •Legacy TCPs silently discard SEG-U >>> • Including the SYN-U that initiates a TCP-U >>> • Three-way handshake between TCP-U and TCP-O times out >>> •This is a feature, not a bug >>> • TCP-U behavior cannot be used as an attack vector against a legacy TCP >>> >>> It's actually more than that: seeing DO=0 immediately tells both endpoints and snooping / meddling middleboxes that the packet format is different that the standard one, WITHOUT having any state and WITHOUT parsing the options. A modest advantage, to be sure, but IMO real nonetheless. >>> >>> I'm agnostic about the coding of option length; either the proposal in the draft or Richard's alternative would be fine. Given the need to make checks that unused bits are zero or at least mask them off, I'm not sure there's a whole lot of difference. >>> >>> Respectfully, >>> >>> Mike Heard >>> >>> From: Richard Scheffenegger <rs.ietf@gmx.at> >>> Sent: Thursday, July 23, 2026 6:17 AM >>> To: tcpm@ietf.org >>> Subject: [tcpm] Re: Fw: New Version Notification for draft-bonica-tcpm-extended-options-04.txt >>> After the discussions in todays meetings, I propose the following: >>> a) rename the draft to something with a higher hamming distance in the TLA to EDO (extended data offset). For example Long Authentication Offset (LAO). This will help to reduce confusion when discussing these two different approaches. >>> b) since this approach has a very clearly described applicability (BGP/TCP authentication), clearly descibing that and NOT claim general applicability for any TCP application can also help focus the work. This also would allow avoiding intricate discussion around middlebox behavor, when the expected use case is dominated by back-to-back connectivity (TTL 1 or 2). >>> From a protocol design perspective, with the 32bit field (1 octet for option-length>>2, 3 reserved).. Is the bit-shifting really needed? Also, a byte in the option value of 0 is "EOL" (end-of-list), putting that in the first octet may have some merit. >>> E.g. >>> +------------------+--------------+-----------------+--------------=2D--+ >>> | must be zero | Reserved | 16 bit length without shift| >>> Am 26.06.2026 um 13:50 schrieb Bonica, Ron: >>> Please review and comment >>> From: internet-drafts@ietf.org <internet-drafts@ietf.org> >>> Sent: Friday, June 26, 2026 7:45 AM >>> To: Bonica, Ron <ronald.bonica@hpe.com>; Boni=a, Ron <ronald.bonica@hpe.com>; Tony Li >>> Subject: New Version Notification for draft-bonica-tcpm-extended-options-04.txt >>> A new version of Internet-Draft draft-bonica-tcpm-extended-options-04.txt has >>> been successfully submitted by Ron Bonica and posted to the >>> IETF repository. >>> >>> Name: draft-bonica-tcpm-extended-options Revision: 04 >>> Title: TCP Extended Options >>> Date: 2026-06-26 >>> Group: Individual Submission >>> Pages: 13 >>> URL: https=//www.ietf.org/archive/id/draft-bonica-tcpm-extended-options-04.txt >>> Status: https=//datatracker.ietf.org/doc/draft-bonica-tcpm-extended-options/=br> HTML: https=//www.ietf.org/archive/id/draft-bonica-tcpm-extended-options-04.html<=span> >>> HTMLized: https=//datatracker.ietf.org/doc/html/draft-bonica-tcpm-extended-options >>> Diff: https=//author-tools.ietf.org/iddiff?url2=draft-bonica-tcpm-extended-options-=4 >>> >>> Abstract: >>> >>> A TCP header accommodates only 40 octets of TCP optio=s. In the >>> future, applications may require more than 40 octets =f TCP options. >>> For example, an application may require a TCP Authent=cation Option >>> (TCP-AO) with a 384-bit MAC. This option requir=s 52 octets of >>> option space. >>> >>> This document defines a TCP extension that accommodat=s up to 1,016 >>> octets of TCP options while maintaining backward compatibility with >>> legacy implementations. >>> >>> >>> >>> The IETF Secretariat >>> >>> >>> >>> _____________________________=_________________ >>> tcpm mailing list -- tcpm@ietf.org >>> To unsubscribe send an email to tcpm-leave@ietf.org >>> >>> >>> >>> >>> This electronic message contains information generated by the USDA solely for the intended recipients. Any unauthorized interception of this message or the use or disclosure of the information it contains may violate the law and subject the violator to civil or criminal penalties. If you believe you have received this message in error, please notify the sender and delete the email immediately. >>> _______________________________________________ >>> tcpm mailing list -- tcpm@ietf.org >>> To unsubscribe send an email to tcpm-leave@ietf.org >> >> _______________________________________________ >> tcpm mailing list -- tcpm@ietf.org >> To unsubscribe send an email to tcpm-leave@ietf.org > > _______________________________________________ > tcpm mailing list -- tcpm@ietf.org > To unsubscribe send an email to tcpm-leave@ietf.org
- [tcpm] Fw: New Version Notification for draft-bon… Bonica, Ron
- [tcpm] Re: Fw: New Version Notification for draft… C. M. Heard
- [tcpm] Re: Fw: New Version Notification for draft… Joseph Touch
- [tcpm] Re: Fw: New Version Notification for draft… Richard Scheffenegger
- [tcpm] Re: Fw: New Version Notification for draft… Tony Li
- [tcpm] Re: Fw: New Version Notification for draft… C. M. Heard
- [tcpm] Re: New Version Notification for draft-bon… Joe Touch
- [tcpm] Re: New Version Notification for draft-bon… C. M. Heard
- [tcpm] Re: New Version Notification for draft-bon… Joe Touch
- [tcpm] Re: New Version Notification for draft-bon… C. M. Heard
- [tcpm] Re: New Version Notification for draft-bon… Joseph Touch
- [tcpm] Re: New Version Notification for draft-bon… Joseph Touch
- [tcpm] Re: New Version Notification for draft-bon… Michael Tuexen
- [tcpm] Re: New Version Notification for draft-bon… Joe Touch
- [tcpm] Re: New Version Notification for draft-bon… Michael Tuexen
- [tcpm] Re: New Version Notification for draft-bon… Joe Touch
- [tcpm] Re: New Version Notification for draft-bon… Bonica, Ron
- [tcpm] Re: New Version Notification for draft-bon… C. M. Heard
- [tcpm] Re: New Version Notification for draft-bon… Michael Tuexen
- [tcpm] Re: New Version Notification for draft-bon… Joe Touch
- [tcpm] Re: New Version Notification for draft-bon… Bonica, Ron
- [tcpm] Re: New Version Notification for draft-bon… Joe Touch
- [tcpm] Re: New Version Notification for draft-bon… Bonica, Ron
- [tcpm] TCP Option Length Bit Shifting (was: New V… Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting Richard Scheffenegger
- [tcpm] Re: TCP Option Length Bit Shifting Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting Michael Tuexen
- [tcpm] Re: TCP Option Length Bit Shifting Richard Scheffenegger
- [tcpm] Re: TCP Option Length Bit Shifting Michael Tuexen
- [tcpm] Re: TCP Option Length Bit Shifting Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting Michael Tuexen
- [tcpm] Re: TCP Option Length Bit Shifting Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting Michael Tuexen
- [tcpm] Re: TCP Option Length Bit Shifting Richard Scheffenegger
- [tcpm] Re: TCP Option Length Bit Shifting Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting C. M. Heard
- [tcpm] Re: TCP Option Length Bit Shifting Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting Michael Tuexen
- [tcpm] Re: TCP Option Length Bit Shifting C. M. Heard
- [tcpm] Re: TCP Option Length Bit Shifting Michael Tuexen
- [tcpm] Re: TCP Option Length Bit Shifting Bonica, Ron
- [tcpm] Re: TCP Option Length Bit Shifting C. M. Heard
- [tcpm] Re: TCP Option Length Bit Shifting Richard Scheffenegger
- [tcpm] Re: TCP Option Length Bit Shifting (was: N… Michael Tuexen