Re: [xrblock] WGLC for draft-ietf-xrblock-rtcp-xr-loss-conceal-05

[Kevin Gross <[e-mail]>]

Without doing more engineering, I can't say for sure.

I think you either need to articulate a better justification than you have
given Colin for sticking with millisecond units or do the comprehensive
engineering required to improve the proposal.

Kevin Gross
[phone]
Media Network Consultant
AVA Networks - www.AVAnw.com <http://www.avanw.com/>, www.X192.org


On Wed, Jul 3, 2013 at 12:09 AM, Qin Wu <[e-mail]> wrote:

>  I am okay to change unit for threshold and reporting in the concealment
> seconds block.****
>
> What about loss concealment block? It looks you suggest to change 16 bit “Mean
> Playout Interrupt Size  ”****
>
> And “Playout Interrupt Count  ” into 32bit? Are I right?****
>
> ** **
>
> Regards!****
>
> -Qin****
>
> *From:* [e-mail] [mailto:[e-mail]] *On
> Behalf Of *Kevin Gross
> *Sent:* Tuesday, July 02, 2013 11:35 PM
> *To:* Qin Wu
> *Cc:* xrblock-chairs; Colin Perkins; xrblock
>
> *Subject:* Re: [xrblock] WGLC for
> draft-ietf-xrblock-rtcp-xr-loss-conceal-05****
>
>  ** **
>
> I'm not suggesting a change to how SCS is measured. We would continue to
> use a 1 second measurement window. I'm proposing a change to the units for
> threshold (to allow a wider range) and the units used for reporting (to
> allow greater precision).****
>
> ** **
>
> Using RTP timestamp units does require more care in sizing fields due to
> the different RTP clock rates used for different types of streams. The most
> demanding case I am aware of is RFC 3497 which uses a 148.5 MHz RTP clock.
> 16-bits is definitely not enough at this rate and 32-bits can even be
> insufficient to the extent that RFC 3497 breaks some basic RTP stuff and
> perhaps should be considered an outlier.****
>
> ** **
>
> On the other hand, although there's a lot of precedent, I believe we need
> to be working away from time representations limited to 1 ms resolution. I
> don't often see a valid justification for arbitrarily limiting resolution
> and there are emerging cases where the additional precision is significant
> and useful. I still intend to write an I-D with timestamping
> recommendations but need to get my existing drafts (clksrc, leap-second)
> through the system before I can take this on. In the meantime, I will
> continue to bring this up in draft comments.****
>
>
> ****
>
> Kevin Gross****
>
> [phone]****
>
> Media Network Consultant****
>
> AVA Networks - www.AVAnw.com <http://www.avanw.com/>, www.X192.org****
>
> ** **
>
> On Mon, Jul 1, 2013 at 8:22 PM, Qin Wu <[e-mail]> wrote:****
>
> Hi, Kevin:****
>
>  ****
>
> I have two difficulties to change measurement unit.****
>
> draft-ietf-xrblock-rtcp-xr-loss-conceal-05 defines two XR metrics
> blocks,one is loss concealment block, the other is concealment seconds
> block.****
>
> I think changing unit from milliseconds (ms) to RTP timestamp unit is more
> applicable to Loss Concealment Block than concealment seconds Block.****
>
>  ****
>
> In the concealment seconds block, you are right, SCS threshold is better
> represented as a fraction 0/255 to 255/255,e.g., 5% of 1 seconds is 50ms.*
> ***
>
> However concealment seconds block is based on successive one second
> intervals as declared by a local clock and used to report the number of xxx
> seconds that occur. See 2nd paragraph of section 4, it said:****
>
> “****
>
>    The following metrics are based on successive one second intervals as**
> **
>
>    declared by a local clock.  This local clock does NOT need to be****
>
>    synchronized to any external time reference.  The starting time of****
>
>    this clock is unspecified.  Note that this implies that the same loss**
> **
>
>    pattern could result in slightly different count values, depending on**
> **
>
>    where the losses occur relative to the particular one-second****
>
>    demarcation points.  For example, two loss events occurring 50ms****
>
>    apart could result in either one concealed second or two, depending****
>
>    on the particular 1000 ms boundaries used.****
>
>  ****
>
> ”****
>
> So are you suggesting concealment seconds not following 1 seond intervals
> decleared by the local clock?****
>
>  ****
>
> In the loss concealment blocks, “Mean Playout Interrupt Size” is defined
> as 16bit field and uses “ms” as unit however****
>
> “On-time Playout Duration”,” Loss Concealment Duration”,” Buffer
> Adjustment Concealment Duration” are all defined ****
>
> As 32bit fields and also use “ms” as unit. So if we change 32 bit fields
> unit from “ms” into RTP timestamp unit,****
>
> How about 16 bit “Mean Playout Interrupt Size” field?****
>
>  ****
>
> Regards!****
>
> -Qin****
>
>  ****
>
> *From:* Kevin Gross [mailto:[e-mail]]
> *Sent:* Sunday, June 30, 2013 11:24 PM
> *To:* Qin Wu
> *Cc:* Colin Perkins; xrblock-chairs; xrblock****
>
>
> *Subject:* Re: [xrblock] WGLC for
> draft-ietf-xrblock-rtcp-xr-loss-conceal-05****
>
>  ****
>
> I doesn't quite make sense to me.****
>
>  ****
>
> It seems like reporting in RTP timestamp units would make things better
> for both the reporter and reportee. The reporter won't need to maintain a
> separate higher-precision counter and then do a conversion for reporting.
> (Or not implement a higher-precision counter and under report in the
> presence of frequent short events.) The reportee receives higher-precision
> reports.****
>
>  ****
>
> SCS threshold might be better represented as a fraction 0/255 to 255/255.
> This resolves your mixed-units concern and allows a wider range of SCS
> behavior.****
>
>
> ****
>
> Kevin Gross****
>
> [phone]****
>
> Media Network Consultant****
>
> AVA Networks - www.AVAnw.com <http://www.avanw.com/>, www.X192.org****
>
>  ****
>
> On Wed, Jun 19, 2013 at 6:35 PM, Qin Wu <[e-mail]> wrote:****
>
> Thanks!****
>
>
> -----Original Message-----
> From: Colin Perkins [mailto:[e-mail]]****
>
> Sent: Thursday, June 20, 2013 1:20 AM
> To: Qin Wu
> Cc: Shida Schubert; xrblock-chairs; xrblock
> Subject: Re: [xrblock] WGLC for draft-ietf-xrblock-rtcp-xr-loss-conceal-05
>
> Qin,
>
> Makes sense, thanks.
>
> Colin
>
>
>
> On 18 Jun 2013, at 04:20, Qin Wu wrote:
> > Good points. I fully agree.
> > However changing measurement unit from RTP timestamp unit to ms may
> cause measurement unit inconsistency.
> >
> > Taking loss concealment metrics block as an example, both 32 bit On-time
> Playout Duration field and 16 bit Mean Playout Interrupt Size uses the same
> measurement unit (ms), if we only change measurement unit for all 32 bit
> fields that carry loss concealment metrics to RTP timestamp unit, this cause
> > Some fields in the block use "ms" as unit, some field uses RTP timestamp
> unit.
> >
> > Taking concealment seconds metrics as another example, "SCS Threshold"
> field also uses milliseconds as unit.
> >
> > Let me know what you think of this?
> >
> > Regards!
> > -Qin
> > -----Original Message-----
> > From: Colin Perkins [mailto:[e-mail]]
> > Sent: Tuesday, June 18, 2013 6:10 AM
> > To: Qin Wu
> > Cc: Shida Schubert; xrblock-chairs; xrblock
> > Subject: Re: [xrblock] WGLC for
> draft-ietf-xrblock-rtcp-xr-loss-conceal-05
> >
> > Qin,
> >
> > The length of time represented by an audio payload tends to be an
> integer number of milliseconds for frame-based codecs, but can have an
> arbitrary length for sample-based codecs. Using RTP timestamp units might
> allow you to precisely match up the timings, if that matters. Plus, don't
> you have RTP timestamp units, but would need to convert to milliseconds?
> >
> > Colin
> >
> >
> >
> > On 13 Jun 2013, at 06:34, Qin Wu wrote:
> >> Colin,
> >> Yes, you are right.
> >> The length of time represented by audio playload in the RTP packet is
> usually measured using ms.
> >> Another rationale is concealment metrics are just terminal related end
> system metrics and its calculation does not need to rely on RTP timestamp
> in the RTP packet.
> >>
> >> Regards!
> >> -Qin
> >> -----Original Message-----
> >> From: [e-mail] [mailto:[e-mail]] On
> Behalf Of Colin Perkins
> >> Sent: Wednesday, June 12, 2013 12:02 AM
> >> To: Shida Schubert
> >> Cc: xrblock-chairs; xrblock
> >> Subject: Re: [xrblock] WGLC for
> draft-ietf-xrblock-rtcp-xr-loss-conceal-05
> >>
> >> Shida,
> >>
> >> I've read the draft, and believe it's in reasonable shape to progress.
> I would be interested in hearing the authors' rationale for choosing ms as
> the measurement unit rather than RTP timestamp units, however. It would
> seem that there might be an argument for using RTP timestamp units, so the
> reports can exactly line-up with the audio data in the RTP packets.
> >>
> >> Colin
> >>
> >>
> >>
> >> On 29 May 2013, at 17:53, Shida Schubert wrote:
> >>> This is an announcement of a 2 weeks XRBLOCK WG last call on
> >>> "Report Block for Concealment metrics Reporting on Audio Applications"
> >>> prior to requesting publication of the document as a proposed standard.
> >>>
> >>> As per discussion at the last meeting, we are running a second
> >>> WGLC on this draft.
> >>>
> >>> Please send your comments, including nits, to the list by the
> >>>
> >>> 12th of June
> >>>
> >>> If you read the draft and you see no issues, concerns, or nits, please
> >>> express the fact that you have no issue progressing the draft on the
> >>> list as well.
> >>>
> >>> The latest version can be found here:
> >>>
> >>> http://tools.ietf.org/html/draft-ietf-xrblock-rtcp-xr-loss-conceal-05
> >>>
> >>> Regards
> >>>
> >>> Shida as co-chair
> >>> _______________________________________________
> >>> xrblock mailing list
> >>> [e-mail]
> >>> https://www.ietf.org/mailman/listinfo/xrblock
> >>
> >>
> >>
> >> --
> >> Colin Perkins
> >> http://csperkins.org/
>
> _______________________________________________
> xrblock mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/xrblock****
>
>  ****
>
> ** **
>