Re: [MMUSIC] WGLC for draft-ietf-mmusic-data-channel-sdpneg

[Paul Kyzivat <[e-mail]>]

Flemming,

On 1/28/16 9:19 PM, Flemming Andreasen wrote:
> Apologies for joining in late here, but I would like to go back to the
> beginning of this thread to understand better why we believe there is a
> need for attributes to specifically define whether they can be used with
> the "dcsa" attribute.
>
> The way I see it, the "dcsa" attribute is simply an encapsulating
> attribute that specifies attributes for the data channel sub-protocol.
> What I believe we need here is a set of clear procedures that explain
> how you take those encapsulated procedures, "decapsulate" them and
> negotiate based on the resulting set of parameters (which may need to
> include a few parameters from the media description containing them).
> This is similar to the approach we took with SDP Capability Negotiation
> (RFC 5939). It has the benefit of specifying a generic algorithm for
> dealing with these, and above and beyond that, it's up to the entity
> generating the SDP to include a set of parameters that make sense for
> the given sub-protocol. This is similar to what we do with normal SDP
> (i.e. don't specify all the valid combinations) and avoids the rather
> painful approach that we have with mux-attributes.

Originally there were two distinct scopes in which attributes could 
appear: session level and media level. Each attribute needs to specify 
at which of those levels it is valid, and the semantics in each scope. 
IANA registries were defined for the attributes and identify which 
scope(s) are applicable to each.

Later we added source level attributes as another scope. And IANA was 
extended to record that scope as well.

The introduction of dcsa clearly adds another scope, at least as 
distinct as source level. If it is important to put scope in the 
registry, then this one belongs there as much as any other.

We *could* rethink this, and reduce the information put in IANA for 
attributes. It would be sufficient for the IANA registry to give the 
attribute name, and a reference to the document where it is defined. (As 
long as we are assured that the information is present in the referenced 
document. It may be that this is not the case for some older attributes.)

But we should be consistent about this.

	Thanks,
	Paul

> Thanks
>
> -- Flemming
>
>
>
> On 12/11/15 10:25 AM, Paul Kyzivat wrote:
>> On 12/10/15 7:25 PM, DRAGE, Keith (Keith) wrote:
>>> I agree with Christian.
>>>
>>> But Paul and myself have had this discussion before. He seems
>>> convinced we can essentially put stuff in the IANA registry that
>>> needs to be in an RFC.
>>
>> Keith - yes, we continue to disagree. But not for the reason you say.
>> I do not, and have never, proposed that things be put into the
>> registry *instead* of into a document.
>>
>> But I do believe that registries are important as an index/discovery
>> tool. You cannot expect people to be aware of every pertinent RFC that
>> may have been published that pertains to something of interest. If I
>> have an interop problem with a new peer, and see something being used
>> that I am not familiar with, then I need a way discover the specs that
>> might pertain to that. The IANA registries serve that function.
>>
>> (There are *some* cases where the registries are more important. When
>> there is something where the bar for extension is low, and no public
>> document is required, then the registry becomes only public place to
>> learn about it.)
>>
>>> It is probably worth others (such as the WG chairs) weighing on this
>>> discussion.
>>
>> Agreed.
>>
>>     Thanks,
>>     Paul
>>
>>> Keith
>>>
>>> -----Original Message-----
>>> From: mmusic [mailto:[e-mail]] On Behalf Of Christian
>>> Groves
>>> Sent: 11 December 2015 00:21
>>> To: [e-mail]
>>> Subject: Re: [MMUSIC] WGLC for draft-ietf-mmusic-data-channel-sdpneg
>>>
>>> Hello Paul,
>>>
>>> On 10/12/2015 3:47 PM, Paul Kyzivat wrote:
>>>> On 12/9/15 11:04 PM, Christian Groves wrote:
>>>>> Hello Juergen and Paul,
>>>>>
>>>>> The proposed text by Juergen is OK.
>>>>>
>>>>> Paul: Is an IANA registry the right place to capture the nuances of
>>>>> what you're describing below? I agree that when people are defining
>>>>> new attributes that they should consider their use with data
>>>>> channels, but shouldn't that be documented in an RFC rather than
>>>>> the IANA registry?
>>>>> People need to read the RFC rather than simply relying on the
>>>>> registry.
>>>>
>>>> It shouldn't *only* be in the IANA registry. The registry serves as an
>>>> index getting you to the right place, and helping you figure out what
>>>> you need to read.
>>> [CNG] This is what I was getting at. It not just a registry update
>>> we're talking about there's a much more significant document
>>> explaining the use of the attributes with dcsa.
>>>>
>>>> Won't it be confusing if the registry covers usage of attributes a
>>>> session, media, and source level but doesn't cover their use in dcsa?
>>> [CNG] The registry is a short cut. Even if the registry mentions that
>>> its a source attribute, a person still needs to read the RFC to see
>>> how it works. I see that people want to use an attribute for its
>>> functionality not because of what the IANA reg says.
>>>>
>>>> If this isn't needed for dcsa, then by the same logic we ought to
>>>> strip out that other information, and reduce the table to contain only
>>>> references to documents.
>>>>
>>>>> Also isn't the consequence of what you're suggesting is that another
>>>>> draft similar to draft-ietf-mmusic-sdp-mux-attributes would have to
>>>>> be completed in order to populate the IANA registry field that you're
>>>>> proposing?
>>>>
>>>> Perhaps. We'd have to discuss the best way. It may be that most of the
>>>> existing attributes can be covered with a couple of default rules, and
>>>> maybe a few exceptions.
>>> [CNG] Still the analysis of each attribute would need to be done and
>>> documented to get to the few rules. I not completely against doing
>>> this, I'm just keen to get this work finished and this will just
>>> delay things.
>>>
>>> Regards, Christian
>>>
>>>>
>>>>      Thanks,
>>>>      Paul
>>>>
>>>>> Regards, Christian
>>>>>
>>>>> On 10/12/2015 2:43 AM, Paul Kyzivat wrote:
>>>>>> Juergen and others,
>>>>>>
>>>>>> What you have below is a good start.
>>>>>>
>>>>>> I still think we need to at least specify how dcsa impacts IANA
>>>>>> registration of attributes.
>>>>>>
>>>>>> The attribute registration is (going to be) updated so that there is
>>>>>> one table containing all attributes, with fields indicating for each
>>>>>> attribute whether it can be used at session level, media level, or
>>>>>> source level.
>>>>>>
>>>>>> Now consider an attribute usable at media and/or source level:
>>>>>>
>>>>>> It *might* *also* be usable with dcsa, or it might not. That will at
>>>>>> least in part be determined by whether the protocol that it applies
>>>>>> to can run over a data channel.
>>>>>>
>>>>>> There was a suggestion (Keith?) that media level attributes be
>>>>>> implicitly allowed with dcsa if the protocol they apply to is
>>>>>> defined over a data channel. I find that plausible. But then the
>>>>>> instructions for registration ought to say that. Also, if there can
>>>>>> be exceptions (attributes that *can't* be used with dcsa even when
>>>>>> the protocol is defined over data channel) then there should be a
>>>>>> way to register that.
>>>>>>
>>>>>> Source level attributes add another complication. They only apply
>>>>>> over RTP. For now, RTP isn't defined over a data channel. (Though in
>>>>>> principle it *could* be.) If it were, then I suppose the ussage
>>>>>> would have to be 'a=dcsa nn source ...'. I don't think we need to
>>>>>> deal with that right now.
>>>>>>
>>>>>> And then there is the possibility of new attributes being defined
>>>>>> that can be used *only* with dcsa, not directly at media level:
>>>>>>
>>>>>> If this is because the protocol is only defined to run over a data
>>>>>> channel, then it could possibly be registered as a media level
>>>>>> attribute (which is then allowed by default with dcsa). If the
>>>>>> protocol were ever defined over another transport the sdp would be
>>>>>> ready for it.
>>>>>>
>>>>>> But if, for some reason, the protocol can run over a data channel
>>>>>> and also over some other transport, but for some reason a particular
>>>>>> attribute only applies with data channels, or only applies without
>>>>>> data channels, then there ought to be a way to record that in the
>>>>>> registry.
>>>>>>
>>>>>> The worst case is for every spec that defines an attribute to pick
>>>>>> its own way to specify all this, and for there to be no hint in the
>>>>>> iana registry.
>>>>>>
>>>>>> IMO it is still better for the iana registry to be given another
>>>>>> field to indicate applicability with data channels.
>>>>>>
>>>>>>      Thanks,
>>>>>>      Paul
>>>>>>
>>>>>> On 12/9/15 8:24 AM, Juergen Stoetzer-Bradler wrote:
>>>>>>> Paul, Christian, Keith,
>>>>>>>
>>>>>>> Would propose to add following two new paragraphs to the end of
>>>>>>> current section 5.1.2.1 (dcsa Attribute) in order to more clearly
>>>>>>> describe the relationship of a subprotocol's attribute and the
>>>>>>> subprotocol's transport protocol, especially data channel.
>>>>>>>
>>>>>>>
>>>>>>>      It is assumed that in general the usages of subprotocol related
>>>>>>> media
>>>>>>>      level attributes are independent from the subprotocol's
>>>>>>> transport
>>>>>>>      protocol.  Such transport protocol independent subprotocol
>>>>>>> related
>>>>>>>      attributes are used in the same way as defined in the original
>>>>>>>      subprotocol specification, also if the subprotocol is
>>>>>>> transported
>>>>>>>      over a data channel and if the attribute is correspondingly
>>>>>>> embedded
>>>>>>>      in a "a=dcsa" attribute.
>>>>>>>
>>>>>>>      There may be cases, where the usage of a subprotocol related
>>>>>>> media
>>>>>>>      level attribute depends on the subprotocol's transport
>>>>>>> protocol.  In
>>>>>>>      such cases the subprotocol related usage of the attribute is
>>>>>>> expected
>>>>>>>      to be described for the data channel transport, e.g. in an
>>>>>>> extension
>>>>>>>      of the original subprotocol specification.
>>>>>>>
>>>>>>>
>>>>>>> Section 5.1.2 already explicitly mentions that (subprotocol
>>>>>>> related) media level attributes may follow a data channel
>>>>>>> declaration (the a=dcmap line). Thus the restriction to subprotocol
>>>>>>> related media level attributes only  seems to be covered already.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Juergen
>>>>>>>
>>>>>>>
>>>>>>> On 01.12.2015 02:12, DRAGE, Keith (Keith) wrote:
>>>>>>>> I think we need to be a little bit careful here.
>>>>>>>>
>>>>>>>> In a large number of cases, the use of the attribute will be
>>>>>>>> unaltered by the transport that is being used, and I do not think
>>>>>>>> we should have a need of saying anything if this is the case - it
>>>>>>>> should be the default.
>>>>>>>>
>>>>>>>> So for example for MSRP, there are a number of attributes
>>>>>>>> associated with MSRP - for file transfer and so on, and as far as
>>>>>>>> I can tell, they are used when MSRP is sent over a data channel in
>>>>>>>> exactly the same way as when MSRP is sent over TCP or SCTP. We
>>>>>>>> don't write special rules for TCP, so why write special rules for
>>>>>>>> datachannel.
>>>>>>>>
>>>>>>>> We do need to identify how we document any special cases, and yes
>>>>>>>> when we find them, drafts like the msrp draft are the way to do
>>>>>>>> that.
>>>>>>>>
>>>>>>>> Additionally, if we find when defining a new attribute, that it is
>>>>>>>> inappropriate for datachannel usage in some way, then we can write
>>>>>>>> that in the defining specification for the new attribute (although
>>>>>>>> I suspect in this case it will be inappropriate to other
>>>>>>>> transports than just datachannel).
>>>>>>>>
>>>>>>>> I do assume this entire discussion is limited to attributes that
>>>>>>>> can be applied to media lines. We might find it useful to say that
>>>>>>>> in the general draft, if we have not done so already.
>>>>>>>>
>>>>>>>> Regards
>>>>>>>>
>>>>>>>> Keith
>>>>>>>>
>>>>>>>> -----Original Message-----
>>>>>>>> From: mmusic [mailto:[e-mail]] On Behalf Of
>>>>>>>> Christian Groves
>>>>>>>> Sent: 01 December 2015 00:53
>>>>>>>> To: [e-mail]
>>>>>>>> Subject: Re: [MMUSIC] WGLC for
>>>>>>>> draft-ietf-mmusic-data-channel-sdpneg
>>>>>>>>
>>>>>>>> Hello Paul,
>>>>>>>>
>>>>>>>> I actually meant to post the reply to the list, sorry for the
>>>>>>>> confusion.
>>>>>>>>
>>>>>>>> I do understand your concerns. For multiplexing it makes sense to
>>>>>>>> check all the existing attributes to see whether they're
>>>>>>>> applicable due to the fact that quite alot may be used when
>>>>>>>> multiplexing occurs.
>>>>>>>> I think the data channel usage is quite a bit more narrow in scope.
>>>>>>>>
>>>>>>>> Whilst this draft does introduce a general purpose mechanism the
>>>>>>>> actual use of the dcsa is being defined by other drafts. E.g.
>>>>>>>> draft-ietf-mmusic-msrp-usage-data-channel on MSRP discusses what
>>>>>>>> happens when both a=setup and a=dcsa:setup are present. msrp-cema
>>>>>>>> is another one discussed, etc.
>>>>>>>>
>>>>>>>> I think you need to understand the application protocol being
>>>>>>>> carried in a data channel before you can say whether particular
>>>>>>>> attributes should be carried in dcsa and what that actually means.
>>>>>>>> Going through the IANA registry wouldn't be trivial and unless the
>>>>>>>> attributes are put in context of an application protocol I don't
>>>>>>>> think it will be terribly helpful.
>>>>>>>>
>>>>>>>> draft-ietf-mmusic-data-channel-sdpneg in cl.5.1.2.1 does indicate
>>>>>>>> that the dsca parameters follow the procedures defined in the
>>>>>>>> subprotocol specific documents. Maybe we could strengthen this to
>>>>>>>> indicate that the use of dsca by an application protocol should be
>>>>>>>> documented in a specification?
>>>>>>>>
>>>>>>>>
>>>>>>>> Regards, Christian
>>>>>>>>
>>>>>>>> On 1/12/2015 11:30 AM, Paul Kyzivat wrote:
>>>>>>>>> Christian,
>>>>>>>>>
>>>>>>>>> On 11/30/15 6:45 PM, Christian Groves wrote:
>>>>>>>>>> Is this level of detail really needed? There are several drafts
>>>>>>>>>> for CLUE, MSRP and BFCP related to the use of this sdpneg
>>>>>>>>>> mechanism.
>>>>>>>>>> E.g.
>>>>>>>>>> CLUE doesn't use dcsa and MSRP provides guidelines on the use of
>>>>>>>>>> dcsa.
>>>>>>>>>>
>>>>>>>>>> Can't we just follow the principle that the use of the dcsa
>>>>>>>>>> attribute should be detailed in a specification and leave it at
>>>>>>>>>> that? I'm not sure what we gain by analysing every existing
>>>>>>>>>> attribute and forcing every new attribute to do this analyses
>>>>>>>>>> when the set of possible parameters in likely to be very small.
>>>>>>>>> I guess it is arguable both ways. Originally there was only
>>>>>>>>> session level and media level. That was always supposed to be
>>>>>>>>> identified, but it wasn't always followed.
>>>>>>>>>
>>>>>>>>> Then we got source level. That has complicated the registry and
>>>>>>>>> the usage. I am not convinced that two SDP experts would come up
>>>>>>>>> with the same list of which attributes can be used at source
>>>>>>>>> level.
>>>>>>>>>
>>>>>>>>> The dcsa attribute becomes the 4th "level" for attributes. It at
>>>>>>>>> least makes sense to me that it ought to be managed analogously
>>>>>>>>> to the source attribute. But I agree it doesn't provide the same
>>>>>>>>> degree of confusion as the others. It only makes sense if the
>>>>>>>>> usage the attribute applies to works over a data channel. And
>>>>>>>>> that should be known to whatever is processing the SDP.
>>>>>>>>>
>>>>>>>>> I see this was a private message. Why don't you repost it on the
>>>>>>>>> list and lets see what other people think? (Feel free to repost
>>>>>>>>> this reply if you wish.)
>>>>>>>>>
>>>>>>>>>       Thanks,
>>>>>>>>>       Paul
>>>>>>>>>
>>>>>>>>>> Regards, Christian
>>>>>>>>>>
>>>>>>>>>> On 1/12/2015 2:36 AM, Paul Kyzivat wrote:
>>>>>>>>>>> One thing occurred to me that we may have forgotten about:
>>>>>>>>>>>
>>>>>>>>>>> This draft introduces the dcsa attribute, which allows other
>>>>>>>>>>> attributes to be used in this new scope. It notes that not all
>>>>>>>>>>> attributes are suitable for use this way.
>>>>>>>>>>>
>>>>>>>>>>> ISTM that in the future other attribute definitions should
>>>>>>>>>>> include an indication of whether they can be used with dcsa,
>>>>>>>>>>> and the IANA registration of the attribute ought to include
>>>>>>>>>>> this. And then there is a need to update the iana registry for
>>>>>>>>>>> all the existing attributes.
>>>>>>>>>>>
>>>>>>>>>>>       Thanks,
>>>>>>>>>>>       Paul
>>>>>>>>>>>
>>>>>>>>>>> On 11/25/15 4:52 AM, Bo Burman wrote:
>>>>>>>>>>>> This is to announce a 4 week Working Group Last Call for
>>>>>>>>>>>>
>>>>>>>>>>>> draft-ietf-mmusic-data-channel-sdpneg-06
>>>>>>>>>>>>
>>>>>>>>>>>> as proposed standard.
>>>>>>>>>>>>
>>>>>>>>>>>> Please review and provide any comments you may have on the
>>>>>>>>>>>> document by Wednesday, December 23, 2015. Comments should be
>>>>>>>>>>>> sent to the document authors and the MMUSIC WG list. If you
>>>>>>>>>>>> review the document but do not have any comments, please send
>>>>>>>>>>>> a note to that effect as well.
>>>>>>>>>>>>
>>>>>>>>>>>> Please also forward this WGLC call to any other interested
>>>>>>>>>>>> parties who may be able to review the draft, asking them to
>>>>>>>>>>>> also direct their comments to the authors and the list as
>>>>>>>>>>>> above.
>>>>>>>>>>>>
>>>>>>>>>>>> Thank you!
>>>>>>>>>>>>
>>>>>>>>>>>>            Bo Burman (MMUSIC co-chair)
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> mmusic mailing list
>>>>>>> [e-mail]
>>>>>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> mmusic mailing list
>>>>>> [e-mail]
>>>>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> mmusic mailing list
>>>>> [e-mail]
>>>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>>>
>>>>
>>>> _______________________________________________
>>>> mmusic mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>>
>>>
>>> _______________________________________________
>>> mmusic mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>
>>> _______________________________________________
>>> mmusic mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>
>>
>> _______________________________________________
>> mmusic mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/mmusic
>> .
>>
>
>