TSC Meeting Notes 2021-06-17
Attendance:
Agenda:
Feature freeze this weekend
3.1 Beta release notes/announcement
Open Source Days agenda/speakers
Standard Optional Attributes
SonarCloud coverage report
Key Dates for Review:
Feature freeze: June 20
3.1 release: July 4
Discussion:
3.1 feature freeze this weekend.
Kimball, how's PR #870 going?
Done a good pass with flipping on the code coverage and scanners.
A failure in the uncompressor for PXR24
API is there but haven’t tested the deep stuff.
I don’t yet know if it’s sufficient for the GPU.
We need docs:
Press release
Release notes
Documentation
Should add a section to the C++ API document
An introduction to how to restructure code to use it. Distilled wisdom.
Kimball, can you start a Google doc we can use to draft this verbiage? Yes.
Larry: Luke from Animal is working on texture benchmarks, a torture test, performance is abysmal. Neither of us have hooked up the new stuff. Hopefully this will be a solution.
Would it be better to just make the changes to OIIO?
Larry: That would save me from doing the same thing.
Initial thing should be fairly straightforward.
Peter: we already have two ways of reading a file, the simplified RGBA version and the full version. This is a third.
Does the new API offer anything for writing? Nothing, just if you want C. It’s faster, or can be faster.
XDR stuff is quite a bit different.
Imath release, with optimized half. We should just proceed.
Cmake option to compile in the tables, since the tables can be too large.
Have made the code safe.
If someone hasn’t updated Imath it will still work.
Open Source Days
Spectral data? Kimbal: the Manuka team said, "Ah, sweet, can we add this and that!” But would require a file change.
Joseph: Standard Optional Attributes:
I’d been trying to get the attributes that are defined in the ACES Container File into the standard list.
Also virtual production ones.
Then pulled the whole thing. There it sat.
Over the past week, we flipped from a state where some of us were pushing to get changes made, to a pull.
John Levin @ ILM: “this is insane, stuff is coming in with inconsistent names.”
John Knoll inquired about standards with Dave Stumpf @ASC
Came up with strategy: bypass SMPTE entirely, refine it with input from various studios.
Not about code review, it’s about semantics.
Can use the PR process to host the discussion. One PR for each attribute/chunk.
Nick: Use a feature branch
Talk to Mike Min at Netflix, has a working group focusing on Camera metadata. He wants metadata in OTIO.
"If we use a term, it has the same semantics."
Nick: There should be a schema. So the data can be carried and not lost.