IETF 79 Codec Minutes: Tuesday, November 9th, 13:00-15:00 Beijing time, Valley Ballroom C Chairs: Mike Knappe Jon Petersen Minute Taker: Magnus Westerlund (thank you!) Audio stream location: http://www.ietf.org/audio/ietf79/ietf79-ch8-tue-noon.mp3 1. Administrivia [10 min] Synopsis: note well, review of agenda, liason status http://www.ietf.org/proceedings/79/slides/codec-3.pdf 2. Guidelines and requirements doc progress [25 min] (draft-ietf-codec-requirements-02 and draft-valin-codec-guidelines-08) Synposis: review of current document status and open issues http://www.ietf.org/proceedings/79/slides/codec-3.pdf Erik Norvell (EN): Guidelines does not express a clear preference for a unencumbered code. Vauge statement that royalty free license equal to no known IPR. Document should state clear Jean-Marc Valin (JMV): Discussed the this issue. Should take all text from the charter into account. Apparently something is missing, will be fixed. Michael Knappe (MK): Creating a separate list to discuss JMV: Differnet issue, little interest for separate list Hassnaa Moustafa (HM): Draft needs consider a solution that doesn't require any licensing. MK: What do you mean? HM: What are the licensing terms MK: We are not going to discuss licensing term. Jon Peterson (JP): We can't really discuss the validity of any claim in the IETF. HM: Support the comments on the mailing list (which?) EN: Questioning on how to trade-off technical quality compared to IPR and Licensing. A clear process would help contributors MK: What answer do you like to have. EN: Clear statement on how much quality improvement a certain encumbrance need to be Koen Voss(KV): IF we had a choice we would prefer work without IPR. Currently we haven't received anything else. Markus Isomaki (MI): What does in practice means to have. Should have discussion if the WG is fine with the current IPR licenses, royalty free ones with conditions. Jean-Marc Valin: I would like to invite the people with issues for the IPR terms. Like to understand what terms are acceptable. MK: Need to find the right venue for that discussion. And understand what the relative contribution. Need to set up a separate list Christina Hoene (CH): Absolutetly not. We need to discuss if we want to have the IPR inside or not. MK: We are allowed to discuss relative merit of proposals. JP: We are not going to remove the right for people to humm according their own view and understanding when it comes to Roni Even (RE): Why can't we humm about if Erik's suggestion should be included. MK: What is EN suggesting? Keith Drage (KD): What is considered a separate discussion? An individual can go to the IPR holders and discuss the terms they would like to see. This can't be done as an organization. EN: Have sent clarifying comment to the list on this. MK: An issue that needs to be resolved. Thought that only the reference code issue needed to be resolved. 3. Codec development status [15 min] ( draft-ietf-codec-opus-00 ) Synopsis: review of codec development status - discussion of the integration of CELT and SILK technologies, along with recent additions (pitch-filtering from Raymond Chen) http://www.ietf.org/proceedings/79/slides/codec-1.pdf 4. Codec design walkthrough (a brief tutorial on the design to-date) [25 min] ( draft-ietf-codec-opus-00 ) Synopsis: separate tutorials of CELT and SILK components http://www.ietf.org/proceedings/79/slides/codec-2.pdf http://www.ietf.org/proceedings/79/slides/codec-4.pdf Erik Norvell: I react on the mentioning of the intention to freeze the byte stream. Shouldn't this be subject to WG consensus? MK: Yes, there will consensus. JMV: Based on shrinking amount of work and comments on the solution, we forecast to not have more improvements in a few months. CH: What enchancements are subject to the IPR? JMV: Some of the improvements are releated to the recent changes in the CELT codec. CH: If it is selective, can you leave it off all the time MK: I meant selective in the frequency. JMV: The gain is selected by the encoder that can 0 if of interest. CH: ? JMV: Instead of doing 2 DCT, we do 8 and integrate them to one common set for the whole period. Lennox: Have you looked at what packet loss does to quality as you have prediction across frames. JMV: For a few exta kb you could skipp that prediction completly. 5. Codec testing status [20 min] http://www.ietf.org/proceedings/79/slides/codec-0.pdf Dan Burnett: Are you considering doing any objective testing with Automatic speech recognizers. As encoding can make MK: Do you have any suggestion for test suites? DB: I will send some suggestion to the list. The main work is charaterization, which happens after codec development. Likely not need massive testing to determine that we support the goals. MK: ? RE: I assume this work when being provided will contain reference codec, test cases etc. MK: Yes RE: Strange that you can freeze the codec before the test plan is ready. MK: Not going to freeze it only have it stable, and possible to unfreeze or make changes to polish things. JMV: Yes, intended to have the codec stable and then polish it if things are found. JL: Is it intended to turn on and off different features to determine how much they contribute? JMV: There are way to many components to do comprohensive testing. Development is an evolution, where new features are added and some removed as they become redundant. Sohel Khan: Are you planning to do tandem testing with other codecs? MK: Under consideration. 6. Timelines, actions and wrap-up [20 min] Synopsis: timeslot was reserved for additional discussion, largely unused as discussion happened throughout the afternoon proceedings