======================================================================== CODEC BOF IETF#76 CODEC Meeting Minutes Thursday, November 12, 2009, 1300-1500 (Afternoon Session I -- Orchid East) Chairs: Eric Burger and Peter Saint-Andre Scribes: Daryl Malas and Enrico Marocco Minutes Editor: Peter Saint-Andre Agenda: Slides: Audio: Chat Log: NOTE: These minutes do not summarize the content of the presentations. Refer to the slides and audio recording for that information. ======================================================================== Introduction and Agenda Bashing Presented by: Eric Burger ======================================================================== IETF NOTE WELL statement reiterated. Jabber scribe and note taker recognized. Logistics noted. Agenda bashing. Eric Burger summarizes BoF results from Stockholm as well as work completed since then. Roni Even: Is the goal of this working group to produce one codec? Eric Burger: Most likely one, but maybe 2 -- this is something for us to discuss. Not going to replicate Study Group 16 and become a codec factory. Eric Burger summarizes the possible approaches... Bob Hinden: In order to do something royalty free implies that we've done an infinite amount of IPR searching and that would be hard to guarantee. Peter Saint-Andre: It is well understood that there is no guarantee. Gregory Lebovitz: A lot of the legal, procedural aspects have been vetted on the mailing list. Stephan Wenger: The only realistic goal is to attempt to standardize something where no IPR disclosures that are not acceptable to the WG have been filed with the IETF. Basavaraj Patil: Are you suggesting that none of the existing codecs are "good", that they are not optimized for the internet, or that they are not royalty free? Eric Burger: There are good codecs and widely implemented codecs and freely-available codecs. The goal is to create one aligning with all three of those features -- it's the intersection of high quality, optimized for the Internet, and royalty free that's empty. ======================================================================== ITU-T Speech and Audio Coding (ITU-T Study Group 16) Presentation Presented by: Yusuke Hiwasaki ======================================================================== Yusuke Hisawaki provides an overview to clarify ITU-T speech/audio coding standardization process, but not an official ITU-T position (refer to the liaison statement). Gregory Lebovitz: Is a sector member an organization that can bring multiple people or an individual? Yusuke Hiwasaki: It can be both. Yusuke Hiwasaki continues presentation, discussing alternative of a focus group... Stephan Wenger: It is my understand that a focus group cannot create a recommendation or a standard, only a report. Is that correct? Yusuke Hiwasaki: Yes. Question relayed from the Jabber room about whether working documents are freely available. Yusuke Hiwasaki: Documents from work at the question level are closed, at least officially. Non-normative comment by Stephan Wenger: Many documents at the "question" level are readily available on an anonymous FTP server. That is not official, published policy, but is often the case. Yusuke Hiwasaki continues presentation, descrbing another alternative: a joint effort, such as the Joint Video Team that produced H.264 between ITU-T SG 16 and ISO/IEC. ======================================================================== Guidelines for proposed codec working group Presented by: Jean-Marc Valin ======================================================================== Jean-Marc Valin provides an overview of the proposed working group procedures, as described in draft-valin-codec-guidelines. Roni Evan: When you say identify requirement: a requirement for a specific codec for a specific application? Roni Evan: At what stage do you go from individual drafts to working group drafts? Eric Burger: That would be a WG decision in accordance with IETF process. Larry Masinter: The first time a WG issues a document, it becomes a WG item. Roni Evan: All of the tasks outlined here would be with reference to individual drafts, no? Eric Burger: That is my assumption, but that is really for the WG to determine, because that is standard IETF practice. Larry Masinter: How would you deal with differences of opinion about the priority of requirements? Russ Housley: Rough consensus and running code. Larry Masinter: But there is a process for getting to rough consensus on things like that, I'm not sure where that is in this process. Eric Burger: This is IETF process. Peter Saint-Andre: Larry, do you think that every WG needs to define processes for reaching consensus, or do you think that there are special requirements in this case? Larry Masinter: It seemed to me that the largest source of dissension would come from requirements for different applications, which would underlie different choices. Having a process where you presume agreement early in the process was fuzzy. Eric Burger: Point taken. Adrian Farrel: On the proposed process, "iteratively improve requirements based on received contributions" -- does this begin at the start of a working group, before contributions are solicited? Jean-Marc Valin: There is no working group at this point, so no contributions can be "improved" yet. But there are already contributions on the table. So there is some overlap here, and that's unavoidable. Colin Perkins: In the past some WGs have assumed that there is some signficance to accepting a draft as a WG item. Concern that there might be a formalized process for determining when an individual draft becomes a working group item. Eric Burger: The plan is to generate one, two, or three documents, but there is no plan at this point to create a formalized process because the plan is to get it done and close the WG. Colin Perkins: I don't think there is a standardized model in the IETF about when to accept an individual draft as a WG item. Jean-Marc Valin continues his presentation... Christian Hoene: What is the definition of "party" with respect to participation / contribution, does that include companies and standardization bodies? Eric Burger: This is the IETF, it's individuals. Jean-Marc Valin: Anyone. Julien Meuric: It looks to me that this typical IETF process, but I would have expected more focus on codec work specifically, such as coordination with other SDOs. Eric Burger: That's not a clarification question, that is coming up in following slides. Ingemar Johansson: Does requirements mean only technical requirements or also non-technical requirements (IPR etc.)? Roni Evan: My understanding is that contributors would need to disclose their code. Eric Burger: This is the difference between IETF and other SDOs, IPR is taken into consideration during the technical work. So, yes, IPR disclosures would be required. Stephan Wenger: I don't agree that this a defining characteristic because in the ITU you can IPR into account. In practice not openly discussed, but in practice it is often discussed. Eric Burger: I promise that we will discuss this. Jean-Marc Valin continues his presentation... Roni Evan: In order to enable evaluation, would someone be required to provide their test results? Jean-Marc Valin: Yes, this is assumed. Roni Evan: It would be helpful to make that clear. It will take you five years to finish this work... Jean-Marc Valin continues his presentation... Colin Perkins: What do you mean by the final codec? I-D that goes to WG Last Call, Proposed Standard RFC, Draft Standard RFC, Internet Standard RFC? Jean-Marc Valin: I am not an expert on IETF process, but I would assume the WG would decide. My personal opinion is that this should happen close in time to WG Last Call. Jean-Marc Valin continues his presentation... Roni Evan: If companies decide to bring their own codec, how do you decide which codec is good? Presumably a result of testing? Jean-MarcValin : Our proposal is to say, bring your codec, and we will work together in an effort to produce something that's even better. Roni Evan: The process is not clear to me. You want to merge them. You need to have some definitions. How do you decide what is better? Eric Burger: So the question is: how do you compare solutions? And the answer may be, this may be one of the first things the working group tries to determine. Jon Peterson: Point of order. The purpose of a working groups is for people to bring their ideas and try to gain consensus. Can we stop asking this question? Xavier Marjou: One of the main goals seems to be to produce a royalty free codec. With this process I see no guarantee that this can be achieved. Peter Saint-Andre: There are no guarantees. Eric Burger: We realize this is a risk. Xavier Marjou: Do you want to create a WG whose main goal is not achievable? Eric Burger: The goal of *guaranteed* royalty free output is not achievable, however there are many existence proofs of royalty free codecs. Julien Meuric: As part of the qualification process, will you consider interoperability with legacy codecs? Jean-Marc Valin: It is not a goal of the working group to have legacy compatibility. Christian Hoene: You want to fulfill all requirements. What do you mean? Isn't that the holy grail? Jean-Marc Valin: We mean the three goals mentioned in the charter. Larry Masinter: Several things in this process seem to be unusual for the IETF: developing code and IPR. Are there precendents in other WGs? Stephan Wenger: An example of doing this was iLBC, which was based on open-source code. Not necessarily a successful example. My personal opinion is that iLBC was not evaluated by people knowledgeable in this field, no interest by people in the room, rubber stamped. Peter Saint-Andre: So the problem was they did not have a real WG such as the one being proposed here. Jean-Francois Mule: Stephan, I disagree that evaluation was not done on iLBC. I have proof I can show you offline. Michael Knappe: There has been a lot of discussion about interworking, how things would work in real environments. Noboru Harada: Regarding bit-exact definitions, what are the plans to ensure quality? Jean-Marc Valin: Our goal will be to ensure quality. The point of the conformance testing tools will be to make sure that the quality is the same. ======================================================================== Charter Discussion Presented by: Peter Saint-Andre ======================================================================== Peter Saint-Andre presenting... Larry Masinter: Why would you want more than one codec? Why does it have an objective to design "a very small number of codecs"? Peter Saint-Andre: Our desire is to come up with one, but in reality, we might end up with several. The objective is open to allowing the potential for more than one. Larry Masinter: So accomodating that as an objective... Peter Saint-Andre: Agreed, maybe the charter needs to state as a goal: "Try to design one". Larry Masinter: Also, "not a rubber stamp" is a negative statement, what is the positive goal that you want to do? Peter Saint-Andre: I think the goal is that people would make contributions and that the group would work together to take the best concepts from each contribution to produce something that is better than any one of the contributions and greater than the sum of its parts. Roni Evan: I'm starting to be happier about what you as saying about the number of codecs, because I think the requirements and charter should say the goal is to develop only one codec. Peter Saint-Andre: Agree this should be changed and/or clarified. Basavaraj Patil: Is the goal for people to bring their codec in and get a "rubber stamp"? Peter Saint-Andre: No rubber stamps. Basavaraj Patil: What is the benchmark and baseline to be considered for publication as an RFC? Peter Saint-Andre: This is the job of the requirements document, based on which the WG will reach consensus. Roni Evan: The problem is there are too many wide-band codecs in the market today. This group should determine one, and recommend it to the community. That would help solve the problem, don't provide options. John Klensin: Concern in that the IETF is not set up to provide conformance testing in order to validate a potential codec. We don't have mechanisms for that here. Colin Perkins: John, we do have such a mechanism. We have interoperability testing before proceeding to Draft Standard. The same process would apply here. The working groups can do this, we did it in AVT. Peter Saint-Andre continues presenting... Eric Burger: Are we missing any goals? Stephan Wenger: (Slide 6) While I do not agree that the goals are reachable, I think these are the right goals. Basavaraj Patil: (Slide 6) Where did these business goals come from? From inside or outside. Peter Saint-Andre: These goals came from previous meetings and the mailing list discussions. Gregory Lebovitz: I have observed that a group of people came into the IETF from the community and put forth a desire to accomplish what's summarized on this slide. Once this set of goals was made public, then there was some agreement and some objections. The community has not come to consensus on the goals, but these represent some of the goals from previous discussions. Peter Saint-Andre: We create WGs for people who want to do the work. We're trying to provide a forum for the people who want to do this work. Noboru Habana: If existing codecs can be presented and meet the criteria, will they be selected for this work (e.g., G.722)? Why should we work on new codecs when we have G.722? Peter Saint-Andre: People coming to the IETF and asking to do this work believe they can do better than existing codecs. Noboru Habana: But we are sure that there is no IPR on G.722. Peter Saint-Andre: There is no point in putting an RFC number on a codec that has already been standardized by the ITU-T. Noboru Habana: What would we do if someone brings an old codec to the WG? Eric Burger: Any existing technology must be at least 20 years old, so we can ensure no IPR issues. I was a naysayer about iLBC, saying that the lawsuits would occur, and they have not occurred. I have been proven wrong about the ability of the community to create, effectively, royalty free work output. Basavaraj Patil: Should the working group rely on an IETF legal "team" to continuously monitor the work to ensure we are not violating IPR? Eric Burger: This was discussed on the mailing list. That would be more expensive than paying royalties. Basavaraj Patil: How are you going to ensure that it will be royalty free? Eric Burger: This is the IETF, we have never strived for perfection. Basavaraj Patil: Why are we doing this in the IETF? Peter Saint-Andre: People have come to the IETF asking to do the work here. Stephan Wenger: If there are a group of people interested in a working group, then it should fall somewhere in the IETF. One point missing: the work must call somewhere within the mission of the IETF. The mission of the IETF is to make the Internet better, and I do not think adding another codec qualifies making the Internet better. I think the goal of this work is to make a codec better for some people on the Internet and a certain group of companies, and not for the Internet as a whole. Koen Vos: I do not agree that G.722 is IPR free. There have been annexes added to G.722 recently that have been essential to make it work over the Internet. Jean-Marc Valin: I don't see how this is different than the rest of the work being done by the IETF. The goal is to develop this work in the same way as all other work in the IETF with regard to IPR. About G.722, I think we can do better. Yusuke Hiwasaki: About G.722, I want to make a clarification: those PLCs (packet-loss concealments) are Appendices, not Annexes. And Appendices are not mandatory part of Recommendations. Also, G.722 has been accepted as a codec for ETSI and it is regarded important. Larry Masinter: When people get together for a standard, everyone has different business goals. I would request the "business goals" are merely a statement of consideration and not a requirement for the charter. Let individuals come up with their own business goals, just provide fair warning about this as a common goal among those who want to participate.. Ingemar Johansson: Are we creating technical work to solve a business problem? Peter Saint-Andre: I think this is semantics, we needed a name for these considerations and didn't want to use the word "legal". Perhaps "other" or "non-technical" considerations is more appropriate. Ingemar Johansson: I do not think the IETF should be trying to do codec work, because there is concern over businesses being able to charge for this. I think this is opening up anyway, and there's no need for the IETF to do this from a technical perspective. Peter Saint-Andre: My goal in assisting with this effort is to help the Jabber open-source community by providing a single codec that can be widely implemented and easily distributed. Stephan Wenger: I do not think the "open-source" community is special. They could become competitors of regular businesses. We should not be working to specifically help them (creating unfair competition). Basavaraj Patil: As far as I know the IETF is working on protocols, and codec expertise does not reside here. The IETF should have told these people to do the work elsewhere. It is incorrect for the IETF to do this kind of work. Joe Hidebrand: (Comments from the Jabber room) - The WG is responsible for balancing considerations and that is the job of the WG. Producing a new codec makes the Internet better if it is adopted, if not then it does not harm. Koen Vos: There have been multiple objections on the mailing list from ITU-T individuals. How does ITU-T think it can be helpful? Eric Burger: Yusuke, could you clarify that, because it's my understanding that in the ITU-T an "objection" has a special status, it's not an objections if just a few people express disagreement. Yusuke Hiwasaki: Correct, objections would be really difficult at ITU-T because that happens at the State level. Xavier Marjou: I'm confused, did you say that you will remove the business goals? Peter Saint-Andre: I said I mislabeled them. Non-technical perhaps, or "other" goals. Christian Hoene: It is good to develop a codec for people who cannot afford the "cost-basis" codecs, the poor people of the world who don't have access to high technology. Peter Saint-Andre: I agree, they are among our stakeholders. Colin Perkins: Regarding IPR rules, there are many other WGs following this process without any issues, they can have their own strong preferences, including for unencumbered technologies. We have lots of existence proofs of other groups working this way, so it should not be a problem. Stephan Wenger: I suggest that we avoid discussions on whether the ITU is appropriate venue and instead focus on whether the IETF is an appropriate venue. Koen Vos: I don't understand why is the ITU-T so worried about the IETF working on this? Concerned about duplication of effort? Yusuke Hiwasaki: Some members of the ITU-T have showed concerns, but I do not think they are speaking as a representatives of the ITU-T. What we are trying to do is to find a way to cooperate, we are willing to help and cooperate in this work. Jean-Marc Valin: I think our views may not be that far apart. The IETF is very willing to work with the ITU-T, and I think forming a working group is a very good method for working with the ITU-T and Study Group 16. This can be the basis of good collaboration. Colin Perkins: There may be people willing to license software if the technology is approved as a standard, but it appears the working group will not accept this because they want a grant up front. Eric Burger: This was addressed on the mailing list. We can assuage those concerns because standard procedure is that IPR is granted upon stndardization, not up front. Christian Hoene: Cooperation with ITU-T is important, but it's not clear how that would happen. Eric Burger: To be worked out. Eric Burger calls questions... Hum: Who here is interested in working on a CODEC in keeping with BCP 79 for use on the Internet? Fair number of hums. Hum forumulation in progress... [Of those who hummed, are the problem statement, charter, and documents clear? Or, do we need to do more work on these items?] Russ Housley: I am concerned with that question. Let's look at the standard questions from RFC 5434. Is there a problem to be solved? Is there a constituency that cares about it? Are there people willing to do the work? Those are the three big questions. I'd like to see people raise their hands. Peter Saint-Andre: Please raise your hand if you are interested in actually working on this within the Internet Standard Process by writing drafts, reviewing drafts, characterization, testing, etc. 20 hands (15 in room/5 Jabber) Peter Saint-Andre: Of those, please raise your hand if you think the IETF is the right place to do this work. 20 hands (14 in room/6 Jabber) Basavaraj Patil: I would like to see a show of hands of people who do not want to see the work done at the IETF... Peter Saint-Andre: Of those who want to do the work, how many do not want to see this work done at the IETF? Zero hands. Gregory Lebovitz: I would like to see, just for the record, a show of hands of the people who don't want to see this work happen at the IETF. We might be able to infer that those people don't want to do the work and don't want to the IETF to do this work. Peter Saint-Andre: Raise your hand if you think that work of this kind should not happen at all at the IETF. 28 hands. Peter Saint-Andre: How many people think this could potentially be done successfully in an interaction between the IETF and the ITU-T? 21 hands. Stephan Wenger: Suggested question: I would like to know how many think this work is achievable / has a reasonable chance of success. Peter Saint-Andre: Raise your hand if you think this work has a reasonable chance of success somewhere (e.g., at IETF or via cooperation between IETF and ITU-T). 30 hands. Peter Saint-Andre: Raise your hand if you think this work does not have a reasonable chance of success anywhere. 8 hands. [audio stream cuts off here!] Gregory Lebovitz: In conclusion, there is a shared desire between the IETF and ITU-T to clearly articulate a practical mechanism to work together. This must be worked out and documented. END