Re: [quicwg/base-drafts] Do we need to define a maximum packet size? (#383)

ianswett <notifications@github.com> Fri, 10 March 2017 14:53 UTC

Date: Fri, 10 Mar 2017 06:53:54 -0800
From: ianswett <notifications@github.com>
To: quicwg/base-drafts <base-drafts@noreply.github.com>
Message-ID: <quicwg/base-drafts/issues/383/285688621@github.com>
In-Reply-To: <quicwg/base-drafts/issues/383@github.com>
References: <quicwg/base-drafts/issues/383@github.com>
Subject: Re: [quicwg/base-drafts] Do we need to define a maximum packet size? (#383)
Mime-Version: 1.0
Content-Type: multipart/alternative; boundary="--==_mimepart_58c2be02c430c_76503fee3d56bc2c29390"; charset="UTF-8"
Content-Transfer-Encoding: 7bit
Precedence: list
Archived-At: <https://mailarchive.ietf.org/arch/msg/quic-issues/jeN2p1jnzzN8QV6C7hC5uMTxLh8>
Cc: Subscribed <subscribed@noreply.github.com>
Reply-To: quic@ietf.org

@martinthomson Dropping packets that are too large doesn't provide much feedback, but that's a failure mode QUIC, at least in Chrome, has to deal with gracefully.  It's much more likely the path drops the packet before it arrives at the server, possibly sending an ICMP packet in the process.

For a while, GQUIC would read packets up to 2000 bytes, but send back an error if the packet was larger than 1460 or something.  This path was almost never hit, but created unnecessary code complexity.

So I'd support no max packet size, with dropping being the way to deal with overly large ones, but possibly we want to require all endpoints support 1500 byte packets as a minimum?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/383#issuecomment-285688621

Re: [quicwg/base-drafts] Do we need to define a m… Martin Thomson
[quicwg/base-drafts] Do we need to define a maxim… Marten Seemann
Re: [quicwg/base-drafts] Do we need to define a m… ianswett
Re: [quicwg/base-drafts] Do we need to define a m… Martin Thomson
Re: [quicwg/base-drafts] Do we need to define a m… Martin Thomson
Re: [quicwg/base-drafts] Do we need to define a m… Igor Lubashev
Re: [quicwg/base-drafts] Do we need to define a m… Marten Seemann
Re: [quicwg/base-drafts] Do we need to define a m… Aron-Schats
Re: [quicwg/base-drafts] Do we need to define a m… ianswett
Re: [quicwg/base-drafts] Do we need to define a m… Aron-Schats
Re: [quicwg/base-drafts] Do we need to define a m… ianswett
Re: [quicwg/base-drafts] Do we need to define a m… Aron-Schats
Re: [quicwg/base-drafts] Do we need to define a m… janaiyengar
Re: [quicwg/base-drafts] Do we need to define a m… Martin Thomson
Re: [quicwg/base-drafts] Do we need to define a m… Igor Lubashev
Re: [quicwg/base-drafts] Do we need to define a m… ianswett
Re: [quicwg/base-drafts] Do we need to define a m… Mark Nottingham
Re: [quicwg/base-drafts] Do we need to define a m… Martin Thomson