AWSF TAC Meeting - February 12, 2020

Video Conference Link

Voting member attendance

Other Attendees

Patrick Hodoul (Autodesk)
JT Nelson (Pasadena Open Source consortium / SoCal Blender group)
Darby Johnston (DJV)
Alex Meddick (Rising Sun Pictures)
Mark Elendt (SideFX)
Jeff Bradley (DreamWorks)

Apologies

Cary Phillips, OpenEXR Representative
Kimball Thurston, Weta Digital Limited

Agenda

Goals for TAC: Year 2 [0:00-0:10]
- Diversity & Inclusion
- GPU for CI
- Windows and Mac
- VFXPlatform 2020
- Runtime testing with Houdini, Maya, Nuke, et al
- New projects
Working groups [0:10-0:30]
- Framework (John)https://github.com/AcademySoftwareFoundation/tac/pull/129
  - CI working group was first example of this model
  - Looking to move forward with working groups on number of subjects, discussed at face to face model
  - Was done mostly ad hoc for CI working group, didn’t have specifically elected lead for the effort, specific timeline or deliverables. Ongoing group with short term goals based on individual needs of the projects. Has gone well for this particular effort.
  - Model suggested by John is more focused and time limited.
  - Gordon: like the idea of specific goals and specific timelines.
  - Daniel: does it make sense to have a process to set the goals of the working group at the TAC level?
  - Gordon: as a general model, the TAC should focus its energies on making sure we tackle the right problems to get desired outcomes, and giving each working group opportunity to tackle issues “out of band”. Different challenges may require slightly different models. Seems to work well internally at Autodesk, so we can focus the time of the wider TAC group.
  - Python 3 example: there may be more focussed experts / people outside the TAC who would be a better fit for that particular activity.
  - Daniel: working groups is a good model for engagement. “What does it mean to be involved in a working group?”. Template lays out model where only the working group lead is officially designated, otherwise activities of the working group are open (to all).
- Proposals
  - Diversity & Inclusion
  - Python 3 (Josh Minor)
    - Pick this up early this year
  - USD
    - Daniel: Cory has had background discussions with Guido and USD team at Pixar
    - Daniel: 2 issues from face to face meeting. Answer request from Pixar to help with support effort on USD message boards to help relieve some of the load on the main USD developers. And effective forum for USD-related discussions.
    - Cory: discussion with the team on how to tackle the support burden, a lot of them related to building USD at individual facilities. Do we expect the chair of the USD team to be part of the USD team at Pixar? Daniel: we don’t have an answer based on our previous experience, it would make sense that it would be someone who is familiar with either the ASWF of USD.
    - Michael: several members of the TAC are responsible for USD implementation at their respective facilities, we probably want to be involved with how to help spread the adoption of USD, there’s a high level opportunity to help promote the adoption of USD at a high level.
    - Cory: can take that feedback back to the USD. Some concern from the USD mailing lists that having discussions on a separate forum might bifurcate some discussions / dilute.
    - Cory: most concrete area where USD team can use help is the support burden, especially related to building.
    - Eric / Cory: a bit of a chicken and egg situation between deciding what the group to do / meet the requirements of the working group document? Sean: shouldn’t the group decide what it wants to do?
    - Daniel: this is a good “test case” for figuring out the scope of what a working group does.
    - Sean: if some amount of deep dive is needed by subject matter experts (who aren’t necessarily TAC members), seeding the working group with those members and task them with defining the scope of the working group might work best.
    - Gordon: agrees with Sean. TAC can define “what is the success of the group”, for instance “remove barriers to success and USD adoption in the industry”, the group can then come up with steps needed to meet this high level goal. “Here are some places we can help, perhaps through CI/CD infrastructure, or additional resources”. Help define what would be missing in current USD offering to help other studio adoption.
    - Eric: second phase besides focus on adoption could be efficiency of effort, make sure that different studios aren’t duplicating efforts related to USD.
    - Daniel: is Cory comfortable with the approach presented above? Cory: yes, but needs to make sure the Pixar USD team is OK. Some discussion on the usd-interest mailing list today. A lot of “community helping community” on that list.
    - Wave: could also have discussions about how we create USD assets, codify / define best practices to using USD. Cory: does that lead to bifurcation from existing mailing list / communication channels?
    - Wave: Notes about USD
    - Gordon: maybe once we have a list of areas we can focus on, at that point we need to have a working model for some aspects that need tight discussion with USD community.
    - Eric: does the Google Group have in person meetings outside of SIGGRAPH? Cory: not outside SIGGRAPH. Working group would be more focussed if it has regular calls for instance.
    - Sean: specific tasks and timelines may conflict with the goal of helping with ongoing support.
    - Daniel: we agree to defer to further discussion to establish scope and goals to first working group goal. Would Cory agree? Yes.
    - JF: should existing USD mailing list be used for working group discussions? Cory: could make sense, maybe as the main communication mechanism, or regular updates to that mailing list.
  - Other areas?
    - Documentation
    - Web site update
Project proposal: OpenShadingLanguage (Larry) [0:30-0:40]
- https://lists.aswf.io/g/tac/message/1196
- No specific discussion on mailing list
- Larry: won’t spend too much time recapping email. Odd that the proposal came day after face to face meeting, but unfortunately couldn’t quite make that deadline for final sign off. Final OK came at 6PM on day of the meeting!
- Endorsement from Imageworks’ part on how well the ASWF has done. Would have been hesitant to make OSL first submission to ASWF, OSL is very important to SPI, but looking at the success of OCIO for instance, excited as to the future of OSL in ASWF.
- GPU centric development is a big focus: project could benefit from more steering by stakeholders.
- Project has hummed along smoothly over the last 10 years, has had a lot of active development, but that doesn’t mean it can’t use the help of being part of ASWF.
- Fit with ASWF couldn’t be better: the majority of ASWF members are using, have contributed to, or are major users of systems that make use of OSL.
- Gordon: thank you Larry!
- Daniel: we have quorum to vote on project incubation status, first step of project adoption to allow the project to start to move into the governance structure.
- JF: how much can you leverage from previous project submissions? Larry: the project has already taken some steps to meet some of the work done on existing projects, for instance already at 88% for the CII badge status. Current plan is to model most of the governance documents on what was done for OCIO with minimal tweaks necessary. Should also help with legal / administrative clearances. Hopefully this project can help to move ASWF forward on some technical aspects, for instance OSL is already using GitHub Actions for CI, which we are looking at for other projects. And definitely need a strong GPU infrastructure, currently no way to test GPU contributions, and have seen some breakage due to lack of GPU enabled testing infrastructure.
- Daniel: is there some internal GPU testing infrastructure? Larry: the internal version embedded in in-house renderer, so that highlights any breaks quickly, but no good / organized way to do it. Also have to make sure that even people working on non-GPU code don’t break the GPU path.
- JT Nelson: Appleseed is now using OSL completely, they dropped their internal shader model, also moving from CPU to GPU. Can check with Appleseed team.
- Larry: one of the next milestones is to assemble the right TSC, that should have not sure the developers, but also stakeholders such as “end clients”. Appleseed could provide someone. TSC meetings will be open, and should open up visibility.
- Daniel: calling for vote, seconded by Wave, unanimous vote YES. Adopted into incubation stage.
Update on candidate projects [0:40-0:50]
- Media player (Darby)
  - Interest in open source technologies for media review in facilities
  - Darby Johnston is the author of the DJV playback application
  - DJV has been around since 2004, cross platform, supports HDR
  - Had lull in development over last couple years, but recently picked up development
  - Transition from Sourceforge to GitHub
  - Looking at ASWF documents as a guideline for some of the work being done on the project
  - Recently released new version, not quite feature parity with previous version
  - Looking at major features such as playlist (OpenTimelineIO integration), audio
  - Looking at feedback from industry as to what features would be useful
  - Looking at opening up the development process and getting more people involved
  - Daniel: asking Henry to articulate some of our interests (from Foundation members) based on discussion from a few meetings ago.
  - Henry: many of the member facilities (and specifically DNeg) have had internal review tools, supporting various formats and capabilities, looking at end-of-life internal tools, evaluating some commercial offerings. Looking at new toolsets, codecs, don’t necessarily feel that all requirements are addressed by commercial offerings. Considering starting a new internal project, would be open source.
  - Darby: is the ASWF adding a media player as an incubation stage process, or this more just an information gathering process?
  - Daniel: some of both, identified as an area of need / interest in ASWF surveys. Can one of the project chairs talk about how developer engagement has worked for ASWF projects?
  - Michael: a big piece of it is getting a handle on who the users are. For OCIO the experience has been that forming a TSC has been beneficial to getting ideas going. Getting a TSC going, even informally has been very valuable.
  - Daniel: what has DJV had issues with, IP issues, closed source codecs, where ASWF could assist?
  - Darby: would be helpful to have legal advice, codecs that don’t have open source versions, what are guidelines for incorporating those types of codecs in an open source project. Building and packaging has been taking too much of his time, so very interested in CI/CD infrastructure. GPU is GPU powered, so very interested in GPU testing infrastructure. Automated packaging infrastructure would help save a lot of time.
  - Brian: OpenCue working on packaging as well, and since working on end user applications, there could be some commonality there.
  - Darby: DJV has been my own project, guided by own wishes and desired, but would be good to build a roadmap once other developers are involved. Not always easy to get structured feedback from DJV users. Python integration identified as useful.
  - Gordon: do you have an idea as to project adoption / contributing to it? Darby: adoption: this is difficult, packages distributed through SourceForce, about 1000 downloads per week, seems to be used worldwide. Email feedback seems very international. Seems to be used a lot by the freelance community and smaller studios, not really in larger studios. Wasn’t trying to compete with commercial offerings.
  - JT: definitely promoted as playback tool in 100% open source pipelines.
  - Darby: would definitely be interested in working with studios with internal tools nearing end of life. Instead of developing tool from scratch, could smaller amount of resources be allocated to a shared project.
- Rez
CI updates [0:50-0:55]
- Working group
- GPU for CI with dynamic instances
  - Will find practical way to move beyond current proof of concept
- Github Actions
  - A lot of interest in moving towards GitHub Actions, will be tested out in some projects
  - GitHub Actions API
  - GitHub CLI for managing PRs and issues
Next meeting: February 26, 2020