[pm-dir] 答复: RFC6390 review of draft-ietf-xrblock-rtcp-xr-decodability

[Qin Wu <[e-mail]>]

Hi,Benoit:
As Alan observed in PM-DIR review, this draft does not define new metrics but refers to metrics that are
clearly defined in a normative reference.
I think we can skip RFC6390 template usage just like PDV draft(RFC6798) did, can't we?

Regards!
-Qin
-----邮件原件-----
发件人: Benoit Claise [mailto:[e-mail]] 
发送时间: 2013年4月10日 21:53
收件人: Qin Wu
抄送: Alan Clark; Gonzalo Camarillo; [e-mail]; Dan (Dan); Shida Schubert; Huangyihong (Rachel); [e-mail]; [e-mail]; Al Morton
主题: Re: RFC6390 review of draft-ietf-xrblock-rtcp-xr-decodability

Hi Qin,

And don't forget the RFC 6390 template usage.

Regards, Benoit
> Hi, Alan:
> Thank for your valuable comments.
> We have updated the draft to incorporate your comments in the new version (-v11).
> The diff is:
> http://www.ietf.org/rfcdiff?url2=draft-ietf-xrblock-rtcp-xr-decodability-11
> Please also see my reply below.
>
> Regards!
> -Qin
> ----- Original Message -----
> From: "Alan Clark" <[e-mail]>
> To: "Gonzalo Camarillo" <[e-mail]>; <[e-mail]>; "Benoit Claise" <[e-mail]>; "Dan (Dan)" <[e-mail]>; "Shida Schubert" <[e-mail]>; <[e-mail]>; <[e-mail]>; <[e-mail]>; <[e-mail]>; "Al Morton" <[e-mail]>
> Sent: Saturday, April 06, 2013 3:10 AM
> Subject: RFC6390 review of draft-ietf-xrblock-rtcp-xr-decodability
>
>
> There are quite a few issues with the draft - I can re-review as soon as
> these are addressed.
>
> Alan Clark
>
>
>
> A. General Comments
>   
> This draft does not define new metrics but refers to metrics that are
> clearly defined in a normative reference.  The normative reference (ETSI
> TR101290) predates RFC6390 however does contain a fairly clear description
> of the metrics with explanation of their usage. It is not recommended that
> this draft redefines the metrics in RFC6390 template form
>
> [Qin]: Exactly.
>
> however there is
> considerable scope for improvement in the clarity of definition of how these
> metrics are used.
>
> [Qin]: Agree.
>   
> B. Applicability Section
>   
> 1.4 Applicability
> Metrics only measure transport stream quality not content stream quality.
> Also the metrics are not defined in this draft but are encodings of the
> metrics defined in ETSI TS 101290.
>   
> Suggest
>   
> ³This block type allows a counts of MPEG Transport Stream quality metrics
> that are measured in accordance with ETSI TR 101290 [ETSI] to be reported by
> an endpoint.  These metrics are useful for identifying bitstream
> packetization and transport stream encoding problems that may affect the
> user¹s perception of a video service delivered over RTP.²
>
> [Qin]: Okay. Your proposed text have been incorporated in (-v11).
>   
> C. Metrics Definitions
>   
> C.1 General
>   
> For clarity the draft should preface the metrics definitions with a general
> explanation of how these metrics relate to ETSI TR101290. TR101290 generally
> defines error events and this draft contains counts of those metrics.
>
>   
> If there are any ³edge² cases where a problem in one measurement interval
> would be reflected in the count in the next measurement interval then this
> should be articulated in the general description and also in the specific
> metric.  For example, a sync byte error is defined as multiple consecutive
> errored sync bytes and if this was reported in an interval it may have
> occurred at the end of the preceding interval or at some time during the
> present interval - hence the description should state that the count may
> reflect a problem in the current or previous interval. This would also be
> the case for PCR errors and even continuity count errors.
>
> [Qin]: Okay, I have added some text in the 2nd paragraph of section 3
> and incorporated your suggested text in (v-11).
>   
> C.2  Sequence numbers
>   
> begin_seq and end_seq
>   
> These definitions simply say ³As defined inS² which requires the reader to
> refer to another document. It is good practice to at least mention what the
> definition refers to and then to include a reference that contains the
> normative definition.
>   
> SoS..
>   
> ³begin_seq: 16 bits
>
> The RTP sequence number corresponding to the start of the measurement
> period, as defined in Section 4.1 of RFC 3611²
>
> [Qin]: Fixed in (-v11).
>   
> C.3 Metrics definitions
> The metrics definitions should contain a firmer statement of what is being
> measured and, if the normative definition is in another standard, then
> clearly state ³as defined in Section X.Y of NNNNN². This applies to all the
> metrics definitions and the example below can be used as a template for
>   
> For example
>   
> Existing language S..
>   
> TS_sync_loss_count: 32 bits
>   
> Number of TS_sync_loss errors in the above sequence number interval.  It is
> calculated based on the occurrence of errors for "TS_sync_loss"parameter
> defined in the section 5.2.1 of [ETSI] (Also see section 5.5.1 of [ETSI]).
>   
> This is very vague language and it is unclear why the ³Also see² reference
> is there.  A better approach is:
>   
> Replacement language (use this format for each of the metrics)
>   
> TS_sync_loss_count: 32 bits
>   
> A count of the number of TS_sync_loss errors that occurred in the above
> sequence number interval.  A TS_sync_loss error occurs when there are two or
> more consecutive incorrect sync bytes within the MPEG TS stream, as defined
> in section 5.2.1 of [ETSI]. This parameter may be used as part of a Service
> Availability calculation, as defined in section 5.5.1 of [ETSI].
>
> [Qin]: Fixed in (-v11).
>   
> C.4 Service Availability
>   
> Following on from the previous comment,  section 5.5.1 of TR101290 describes
> a service availability error as a combination of TS_sync_loss, PAT_error and
> PMT_error whereas draft-ietf-xrblock-rtcp-xr-decodability-10 does not
> contain the PAT and PMT error metrics.  The resolution for this would either
> be to remove the reference to 5.5.1 or to add the metrics required to
> calculate the service availability.
>   
> [Qin]: Agree. I prefer to remove the reference to 5.5.1 since there was consensus in the past WGLC to this draft
> that having a second report block later to cover the other parameters and get inline with concept of RFC6792
> and letting this draft focus on PSI indpendent parameter reporting.
> See details for the WGLC discussion in the following link:
> http://www.ietf.org/mail-archive/web/xrblock/current/msg01032.html
>
> It is recommended that PAT_error , PAT_error_2,  PMT_error and PMT_error_2
> be included as metrics as these ³are² generally present in MPEG Transport
> streams and errors within these can prevent correct decoding of the stream.
>   
> C.5 PCR_error_count
>   
> PCR_error_count is defined twice - the second of these should be
> PCR_accuracy_error_count
>   
>   [Qin]: Good catch and have fixed in (-v11).
>
>
>