W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2014

Re: h2 Connection Preface

From: Simpson, Robby (GE Energy Management) <robby.simpson@ge.com>
Date: Wed, 10 Sep 2014 17:34:25 +0000
To: Martin Thomson <martin.thomson@gmail.com>
CC: "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>
Message-ID: <D03601A1.39FCE%Robby.Simpson@GE.com>
On 9/10/14, 12:01 PM, "Martin Thomson" <martin.thomson@gmail.com> wrote:


>On 9 September 2014 07:10, Simpson, Robby (GE Energy Management)
><robby.simpson@ge.com> wrote:
>> Why is piggybacking of ACKs disallowed?  We could thus save 9 octets,
>> particularly during the connection preface.
>
>
>I missed this earlier.  There was a preference expressed for a clean
>demarcation between settings and acknowledgements.  I don't think it
>was a strong preference, but since the only reason to allow
>piggybacking was to save bytes and that saving was for a relatively
>rare frame type, I think that it was easier to go with the no
>combination option.

Thanks for the background Martin.

My preference would be to allow piggybacking.  I understand it is
relatively rare, but it is possible during the connection preface in most
configurations.  The potential to save 9 octets during the setup of a
connection is particularly important for short-lived connections,
energy-limited applications, etc., particularly for a protocol with a
stated goal of improving latency and performance.  I'm specifically
thinking of servers sending a piggybacked ACK along with their SETTINGS
during the connection preface.

That said, I understand the ship may have sailed on this one.  But I am
curious if such a change is possible.  Those who wish to maintain a clean
demarcation could simply choose not to send piggybacked ACKs.
Received on Wednesday, 10 September 2014 17:34:57 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 30 March 2016 09:57:10 UTC