W3C home > Mailing lists > Public > ietf-http-wg@w3.org > October to December 2012

Re: new version of Flow Control Principles

From: Roberto Peon <grmocg@gmail.com>
Date: Thu, 13 Dec 2012 16:07:12 -0800
Message-ID: <CAP+FsNdCCYmzOiyS=9_OH+8pf7_7L-vsb+BbDh3OubpZfjiDdw@mail.gmail.com>
To: Jonathan Ballard <dzonatas@gmail.com>
Cc: Gabriel Montenegro <Gabriel.Montenegro@microsoft.com>, "draft-montenegro-httpbis-http2-fc-principles@tools.ietf.org" <draft-montenegro-httpbis-http2-fc-principles@tools.ietf.org>, HTTP Working Group <ietf-http-wg@w3.org>
I'm not understanding what you mean by debit. Do you mean revoking some of
the flow-control-receive-window after the fact?
Can you explain?
-=R


On Thu, Dec 13, 2012 at 12:59 PM, Jonathan Ballard <dzonatas@gmail.com>wrote:

> For wireless, an algorithm may provide semantic value. On the wire, that
> value, or flow, is already present. I would like to say, replace SPDY's
> Push/Pull semantic flow with credit and debit accounting. I thought we
> already understood balances and currencies.
>
> One sample is Raspberry PI FTP box that runs on solar power. It appears
> wasteful not to match power units to credit/debit the flow; when, we can
> avoid energy demands spent on extra algorithms that achieve the same. Power
> is sent over USB, so that is orthogonal to power over Ethernet. The
> blackbox in this sample is the batteries. I can think of many explorations
> projects to fine tune such accounting.
>
> My opinion, credit-only numbers lead to more permanent registries as trust
> tokens develop. That means the blackbox converts into greybox, which may
> not be desired at all. That means more resistance to SPDY.
>
>
> On Thursday, December 13, 2012, Gabriel Montenegro wrote:
>
>>  ** **
>>
>> ** **
>>
>> *From:* Jonathan Ballard [mailto:dzonatas@gmail.com]
>> *Ö*****
>>
>> ** **
>>
>> After I read it the question that was left was, "where is the debit
>> rule?" If that question doesn't make sense then maybe "credit" isn't the
>> real number word for flow control.
>>
>> Gab> Not sure I understand the comment. Could this be related to Eliotís
>> comment that the algorithm is clearly in the hands of the receiver? All the
>> sender has to do to debit towards any available credit is to send as much
>> of that credit as he needs to send. The actual over-the-wire(less)
>> transmission dynamics, of course, are in the hands of the underlying TCP
>> implementation (which would presumably heed congestion control rules, but
>> that is opaque to HTTP 2.0).****
>>
>> ** **
>>
>> Thanks,Gabriel****
>>
>
Received on Friday, 14 December 2012 00:07:42 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 14 December 2012 00:07:47 GMT