Re: Bug 25269

On Fri, Jun 6, 2014 at 4:47 PM, Mark Watson <watsonm@netflix.com> wrote:

> Key Ids are generally random 16-byte identifiers. I don't think we can
> assume that a title id can be derived from a key id except by a database
> lookup.
>

If the key IDs are truly random, they should be unique and #2 should work.

>
The spec does not need to assume either solution. If someone wants to use
these mechanisms, they should work with their packager to enable them.

>
> ...Mark
>
>
> On Fri, Jun 6, 2014 at 4:39 PM, Joe Steele <steele@adobe.com> wrote:
>
>> You are right — as a random 16-byte value it *should* be unique. However
>> the CENC spec does not seem to say that *must* be the case. Thus my
>> concern.
>>
>> I am looking at this text (emphasis mine) from ISO/IEC CD 23001-7 2nd
>> Edition:
>> "NOTE: For global uniqueness, a UUID [1] *should *be used for each
>> unique KID /key value pair to prevent
>> duplicate IDs for different keys by independent publishers. Publishers
>> may use the same key value and
>> KID  in more than one track or file according to their rights management
>> intentions."
>>
>> Option #2 only works if the keys are unique for all titles as you point
>> out. This is a catch-22. My assumption is that they are not unique.
>> Option #3 is a non-starter because the CDM needs to be able to attest to
>> the information in the license request and it can’t unless it has all of it.
>>
>> I suggest we add a note in the spec that emphasizes that these must be
>> unique and suggests Option #1 as a good way to do it.
>> That seems like minimal effort. If implementers run into this problem it
>> will point them in the right direction even if they can’t make the
>> suggested change.
>>
>
That sounds good to me. I'll add it to my TODO list.
>
>
>> Joe
>>
>> On Jun 4, 2014, at 10:47 AM, David Dorwin <ddorwin@google.com> wrote:
>>
>> Yes, it is just an array of key IDs.
>>
>> There are a few options to identify the title:
>>
>>    - The key ID(s) could include the title ID in them, allowing it to be
>>    easily extracted on the server. (This forces unique key IDs.)
>>    - The server can look the title up based on the key ID(s). (This
>>    assumes they are unique.)
>>    - The application knows the title and can include this in its request.
>>
>>
>>  Why do you say KID isn't generally unique across streams? That should
>> probably be a best practice for content providers.
>>
>> David
>>
>>
>> On Tue, Jun 3, 2014 at 10:42 AM, Joe Steele <steele@adobe.com> wrote:
>>
>>> https://www.w3.org/Bugs/Public/show_bug.cgi?id=25269
>>>
>>> This bug was closed on Saturday. Rather than just re-open it — maybe you
>>> can give me some clarification.
>>>
>>> Your proposal only includes the KIDs in the data that is sent into the
>>> CDM — correct?
>>>
>>> Given that KID is not generally unique across streams, how are you
>>> planning to support this? I would expect implementations to require a title
>>> identifier to pass to the server in order to identify the title that the
>>> KID is associated with. It sounds like you are proposing that the
>>> application which has this information should send it to the server in some
>>> out of band communication.
>>>
>>> Joe
>>>
>>
>>
>>
>