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

Re: #540: "jumbo" frames

From: Roberto Peon <grmocg@gmail.com>
Date: Wed, 25 Jun 2014 13:22:00 -0700
Message-ID: <CAP+FsNdPSsOEkbcdV84VQyvcmCF9dWXTtb-VbyAQYv2QjkEB8Q@mail.gmail.com>
To: Jason Greene <jason.greene@redhat.com>
Cc: K.Morgan@iaea.org, Patrick McManus <pmcmanus@mozilla.com>, Nicholas Hurley <hurley@todesschaf.org>, Mark Nottingham <mnot@mnot.net>, Poul-Henning Kamp <phk@phk.freebsd.dk>, Willy Tarreau <w@1wt.eu>, Greg Wilkins <gregw@intalio.com>, HTTP Working Group <ietf-http-wg@w3.org>, Martin Dürst <duerst@it.aoyama.ac.jp>
The sender determines how to apportion data into frames. We were seeing
whole files transmitted as a single frame.
And again, in the web use-case this is not going to be a measurable
difference in utilization, since we're likely to be using TLS.
-=R


On Wed, Jun 25, 2014 at 1:15 PM, Jason Greene <jason.greene@redhat.com>
wrote:

>
> On Jun 25, 2014, at 3:06 PM, Roberto Peon <grmocg@gmail.com> wrote:
>
> >
> >
> >
> > On Wed, Jun 25, 2014 at 1:00 PM, <K.Morgan@iaea.org> wrote:
> > On 25 June 2014 21:48, grmocg@gmail.com wrote:
> > > The same is true for any other optional thing under the sun.
> > > It doesn't seem to be a motivating argument for including more
> optional features, though.
> > > -=R
> >
> > That's not the point.  Patrick implied that adding the jumbo frame
> option kills the ability to efficiently mux.  That's simply false.
> >
> > A jumbo frame option requires the implementation to be smart to before
> the protocol becomes effective/useful for web traffic.
> >
> >
> > On Tuesday,22 April 2014 09:00, fielding@gbiv.com said [1]:
> > "Likewise, restricting packet sizes to a small length in order to
> prevent fools from HOL blocking
> >  their own multiplexed channels makes some sense, for browser developers.
> >  However, it actively harms applications of HTTP that are not interested
> in multiplexing
> >  because they only want to transmit a single large data stream.
> >  ... I don't think it makes sense to limit an application-level protocol
> to the worst case.
> >  ....Roy"
> >
> > So let's be straight.  There are legitimate use cases for jumbo frames.
>  You and Patrick just think the world is full of a bunch of what Roy calls
> "fools" (see above) who will HOL block their own multiplexed channels.  Roy
> (and I) don't disagree.  Roy (and I) don't think it makes sense to limit
> the protocol to protect against an unknown number of fools.
> >
> > Keep in mind that the original SPDY had much, much larger max
> framesizes, and I designed it that way in concert with Mike.
> > Implementation experience, however, provided hard data that this
> actually happens, and harms the user.
>
> Happens from where, the browser? Since the browser controls the channel,
> it can simply set whatever frame size it wants.
>
> If it wishes to implement fast downloading of large files, it can simply
> open a new connection using the larger size and limited to no multiplexing.
>
> --
> Jason T. Greene
> WildFly Lead / JBoss EAP Platform Architect
> JBoss, a division of Red Hat
>
>
Received on Wednesday, 25 June 2014 20:22:28 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:14:31 UTC