Re: Validation of Packet Too Big Payload using Echo Request

神明達哉 <jinmei@wide.ad.jp> Thu, 30 January 2020 22:07 UTC

MIME-Version: 1.0
References: <CAOSSMjVzTmCRxLg8k+9t+j0u9iMMWa+T_8P7kVRKGLkiup_jDw@mail.gmail.com> <F9ACCBEA-D009-48BF-B2F6-7179C0DDCB2A@gmail.com>
In-Reply-To: <F9ACCBEA-D009-48BF-B2F6-7179C0DDCB2A@gmail.com>
From: 神明達哉 <jinmei@wide.ad.jp>
Date: Thu, 30 Jan 2020 14:07:38 -0800
Message-ID: <CAJE_bqeFO4hwJZaa_ChiEwP_zxdiaVS3QmoBAV3NFPNcwCWi_g@mail.gmail.com>
Subject: Re: Validation of Packet Too Big Payload using Echo Request
To: Fred Baker <fredbaker.ietf@gmail.com>
Cc: Timothy Winters <twinters@iol.unh.edu>, 6MAN <6man@ietf.org>
Content-Type: text/plain; charset="UTF-8"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ipv6/xJ2d9lrT9MBkTYh-DzRb0Gz-H5s>
Precedence: list

At Mon, 27 Jan 2020 11:39:58 -0800,
Fred Baker <fredbaker.ietf@gmail.com> wrote:

> PTB and PMTU activates in IPv4 applies to streams of packets, not
> sessions. Given that IPv6 operates from the same definitions, I
> don't see why it would be applied to something different in IPv6.

> Now, in fairness, TCP has a way to address PTB/PMTU - it changes the
> MSS in use for a session. UDP doesn't have anything so clear. But
> IMNSHO, that is a deficiency in UDP, not in ICMP. UDP, or the
> process using it, is informed of the limitation. It fails to act on
> it.

We can still fragment the subsequent UDP packets at the known PMTU (I
know everyone hates IP fragmentation but that's a different topic).
In my understanding PMTUD work less well with UDP application than TCP
apps because UDP apps tend to be "connectionless" (e.g. using
non-connected UDP sockets) and the OS's ICMP layer often doesn't like
to adjust its behavior upon receiving a packet too big error when it
cannot find an application that matches the src-dst pair specified in
the ICMP error (which is not the case for connectionless UDP apps).

BTW, RFC3542 provides a way to let UDP apps (even connectionless ones)
know the change of the path MTU that may affect those applications,
although I suspect RFC3542 in general, let alone this particular
option, is not so widely implemented.

Maybe an offtopic for this thread anyway, though.

--
JINMEI, Tatuya

Re: Validation of Packet Too Big Payload using Ec… Fred Baker
Re: Validation of Packet Too Big Payload using Ec… Fernando Gont
Re: Validation of Packet Too Big Payload using Ec… 神明達哉