Re: Submitted HTML5 bugs and an MSE observation



On 3/18/14, 4:18 AM, "Silvia Pfeiffer" <silviapfeiffer1@gmail.com> wrote:

>I am not following this discussion (I don't know enough about MPEg-2),
>so if there are any bugs in the patches that I proposed, can you
>please post actual text change suggestions in the bug? Thanks.

Done. See https://www.w3.org/Bugs/Public/show_bug.cgi?id=24997#c10


Bob
>
>Cheers,
>Silvia.
>
>On Tue, Mar 18, 2014 at 8:03 AM, Bob Lund <B.Lund@cablelabs.com> wrote:
>>
>>
>> On 3/17/14, 2:55 PM, "Clift, Graham" <Graham.Clift@am.sony.com> wrote:
>>
>>>In my understanding the payload_start_indicator is present in the pid
>>>packet for the particular private section but that the  original
>>>expectation was that creation of the texttracks was done right after
>>>deciphering the PMT. Now are you suggesting that the UA doesn't create a
>>>texttrack until there is a private payload? This adds unnecessary
>>>complexity IMHO to the UA implementation without much benefit.
>>
>> I wasn't suggesting anything. I said that I thought the language in
>> question was technically correct. It would be simpler if the UA created
>>a
>> metadata track based solely on the stream_type; no cues would be
>>generated
>> if there were no private sections.
>>
>>>
>>>
>>>Also, I am no sure that the UA should be decoding the private section
>>>table at all according to an ISO spec that is required to be purchased,
>>>i.e. IS 13818-1.
>>
>> Presumably a UA that creates MPEG_2 TS text tracks also plays MPEG-2 TS
>> media and the implementor will have access to the spec.
>>
>>> Why not just cue the complete 188 byte packets (which would include the
>>>payload_start_indicator) as they are received and let the application
>>>decode into relevant data?
>>
>> Possible I guess but the UA is already processing the streams for the
>> media resource.
>>
>>>
>>>Regards
>>>
>>>Graham
>>>
>>>-----Original Message-----
>>>From: Bob Lund [mailto:B.Lund@CableLabs.com]
>>>Sent: Monday, March 17, 2014 12:23 PM
>>>To: Clift, Graham; Silvia Pfeiffer; Ota, Takaaki
>>>Cc: public-inbandtracks@w3.org
>>>Subject: Re: Submitted HTML5 bugs and an MSE observation
>>>
>>>After rereading the proposed text to resolve the bug, I think it is
>>>technically correct. Payload_start_indicator == 1 signifies presence of
>>>private sections. Stream_type == 0x05 by definition contains private
>>>sections, stream_type == 0x80 - 0xFF may contain private sections. So
>>>the
>>>UA, upon seeing 0x80 - 0xFF, could wait for indication of
>>>private_sections before creating the metadata track. On the other hand,
>>>if the UA creates a metadata track for stream_type, say 0xD8, but the
>>>stream doesn't contain private sections there would be no cues
>>>generated.
>>>
>>>Bob
>>>
>>>
>>>On 3/17/14, 1:04 PM, "Bob Lund" <B.Lund@CableLabs.com> wrote:
>>>
>>>>
>>>>
>>>>On 3/17/14, 11:49 AM, "Clift, Graham" <Graham.Clift@am.sony.com> wrote:
>>>>
>>>>>It seems like for TextTrack creation in MPEG-2 TS that detection of
>>>>>the metadata private section is only through the stream_type code
>>>>>present in the PMT and that the payload_start_indicator==1 is only
>>>>>relevant to populating the DataCues that get added to the TextTrack
>>>>>later.
>>>>
>>>>Yes, you©öre right - the payload_start_indicator should not be used to
>>>>determine metadata track; only the stream_type.
>>>>
>>>>> Hence the patch and the wiki page seem incorrect to me when they
>>>>>suggest the payload_start_indicator==1 is used to detect that there is
>>>>>a stream type of kind metadata private section.
>>>>>
>>>>>This also raises another question which is how are the metadata
>>>>>private section table cues are generated. If the private section table
>>>>>is delivered as a payload split across multiple TS packets then
>>>>>presumably the UA must use the payload_start_indicator bit to begin to
>>>>>assemble the payload into the single private section table. However,
>>>>>when does the cue get created unless the UA knows how many packets to
>>>>>expect?
>>>>
>>>>The private_section_length field can be used by the UA to know when
>>>>it©ös received the last packet for an instance of the table.
>>>>
>>>>> My understanding is the UA does not know how many packets to expect
>>>>>so maybe it generates the cue only when a subsequent
>>>>>payload_satart_indicator==1 is detected. However, this could be long
>>>>>past the time when this cue is relevant to the media playback time.
>>>>>
>>>>>Regards
>>>>>
>>>>>Graham Clift
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>-----Original Message-----
>>>>>From: Silvia Pfeiffer [mailto:silviapfeiffer1@gmail.com]
>>>>>Sent: Monday, March 17, 2014 1:21 AM
>>>>>To: Bob Lund
>>>>>Cc: public-inbandtracks@w3.org
>>>>>Subject: Re: Submitted HTML5 bugs and an MSE observation
>>>>>
>>>>>On Fri, Mar 14, 2014 at 5:56 AM, Bob Lund <B.Lund@cablelabs.com>
>>>>>wrote:
>>>>>> Hi CG,
>>>>>>
>>>>>> I have submitted the following HTML5 bugs 24986, 24977, 25005 and
>>>>>> 25044, reflecting the discussion we've had on the mailing list.
>>>>>
>>>>>[HTML editor hat on:]
>>>>>I've provided feedback and proposed patches in each of these bugs.
>>>>>Please review.
>>>>>
>>>>>Regards,
>>>>>Silvia.
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>

Received on Tuesday, 18 March 2014 14:59:08 UTC