Re: [httpstreaming] QoE feedback support

Gunnar Heikkilä <> Wed, 10 November 2010 15:25 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 38FFF3A6A49 for <>; Wed, 10 Nov 2010 07:25:44 -0800 (PST)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -6.299
X-Spam-Status: No, score=-6.299 tagged_above=-999 required=5 tests=[AWL=-0.001, BAYES_00=-2.599, HTML_MESSAGE=0.001, MIME_8BIT_HEADER=0.3, RCVD_IN_DNSWL_MED=-4]
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id G7qyxDebJAp6 for <>; Wed, 10 Nov 2010 07:25:42 -0800 (PST)
Received: from ( []) by (Postfix) with ESMTP id 6F44F3A69F6 for <>; Wed, 10 Nov 2010 07:25:42 -0800 (PST)
X-AuditID: c1b4fb3d-b7b28ae00000135b-d2-4cdab990ef4f
Received: from (Unknown_Domain []) by (Symantec Mail Security) with SMTP id 64.7C.04955.099BADC4; Wed, 10 Nov 2010 16:26:09 +0100 (CET)
Received: from ([]) by ([]) with mapi; Wed, 10 Nov 2010 16:26:08 +0100
From: =?iso-8859-1?Q?Gunnar_Heikkil=E4?= <>
To: "Ali C. Begen (abegen)" <>, David Singer <>
Date: Wed, 10 Nov 2010 16:26:07 +0100
Thread-Topic: [httpstreaming] QoE feedback support
Thread-Index: AcuA55sVlUKpEd3XQBSBesf8npc0JQAAmiGw
Message-ID: <>
References: <> <> <>
In-Reply-To: <>
Accept-Language: en-US
Content-Language: en-US
acceptlanguage: en-US
Content-Type: multipart/alternative; boundary="_000_BCAD297FC0C0D244894589EE45FE8B470FF21A9D8BESESSCMS0364e_"
MIME-Version: 1.0
X-Brightmail-Tracker: AAAAAA==
Cc: Ingemar Johansson S <>, httpstreaming <>
Subject: Re: [httpstreaming] QoE feedback support
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Network based HTTP Streaming discussion list <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Wed, 10 Nov 2010 15:25:44 -0000

The acronym QoE is maybe not the best choice of name, but the concept allows the 3GPP operator to get a better understanding on how the streaming service works for the end client. Some more integrated services (like circuit-switched AMR voice) can typically be handled by measuring things within the 3GPP network itself (for instance codec mode adaptation, frame loss etc.), but other services (like HTTP streaming) have a lot more freedom for the end client to decide how to play out the media.

This is a good thing, as clients can choose different strategies to satisfy the end user, but it also means that it is difficult for the operator to understand if his network works good enough for these type of services. The metrics reported back from the client are typically things which are not possible to measure in the network, but which does affect the media "enjoyment" a lot. For instance, the length of the initial buffering, or any re-buffering occurances etc.

Not rocket science, not perfect, but still information which help the operator to see a more complete picture of his network and the services used...


From: [] On Behalf Of Ali C. Begen (abegen)
Sent: on 10 november 2010 15:57
To: David Singer
Cc: Ingemar Johansson S; httpstreaming
Subject: Re: [httpstreaming] QoE feedback support

Sure but this does not answer my question afaict.


On Nov 10, 2010, at 9:09 PM, "David Singer" <<>> wrote:

On Nov 10, 2010, at 13:52 , Ali C. Begen (abegen) wrote:

> I really find it very unattractive to call this stuff QoE reporting/feedback. Give me one example on what metric would measure my level of pleasure from a streaming service ;)
> -acbegen
>> -----Original Message-----
>> From:<> [<>] On Behalf Of Ingemar Johansson S
>> Sent: Wednesday, November 10, 2010 8:42 PM
>> To: httpstreaming
>> Subject: [httpstreaming] QoE feedback support
>> FYI
>> This document describes how QoE support is proposed for HTTP streaming in 3GPP.
>> <>
>> I suspect that there exist a newer version and/or a better description.

we were actually just discussing this area at a 3G meeting, and I realized that there are really two kinds of metrics it is useful to get from clients:

a) truly Qo Experience, that is, experiential metrics -- the mismatch between the user's expectation, and reality (tune-in delay, media they wanted to see but did not, and so on)
b) operational metrics -- amount of data fetched, buffer sizes maintained, and so on

There are some in the middle, like bandwidth experienced, which are both operational and affect experience.

David Singer
Multimedia and Software Standards, Apple Inc.