Attendance:

  • Cary Phillips
  • Christina Tempelaar-Lietz
  • John Mertic
  • Joseph Goldstone
  • Kimball Thurston
  • Larry Gritz
  • Nick Porcino
  • Peter Hillman
  • Rod Bogart

Discussion:

  • Metadata discussion
    • Regarding conversation on Slack thread
    • Joseph: for Sam - necessary to strip metadata string and convert to float
    • Larry: more the name than the metadata value
    • Joseph: will take a look into this, common mistake to get the attribute name wrong, issue in Arri raw to exr converter
    • Larry: some nuggets to thing about - should there be API support for querying? Recommendation for how to do namespacing?
    • Cary: could say on website : preferred separator is … 
    • Joseph: any notion of deprecation? If something is being renamed, return more than just a simple boolean? Or iterate over all of them?
    • Peter: idea of deprecating should become part of the standard
    • Larry: other areas where we could put out guidance on best practice. Should metadata names with no namespaces be flagged, somewhat enforced? Should you be strongly encouraged into using standard names? What are precepts for being a good citizen in naming convention of attributes.
    • Joseph: prefixed Arri-specific attributes - e.g. com.arri.mirrorshutter. Others weren’t defined in ACES container spec, and put a prefix of “interim” so if you came across “interim” and non-interim attributes with same name, non-interim wins. If generating new files, deprecation lets application writer know not use that name. Challenge if two different applications are adding both old and new versions of that named attribute.
    • Peter: considered adding an API call to hand your header that verifies if you are doing everything right.
    • Larry: ran into a situation like this recently where wrong type was written for an attribute in sample.
    • Larry: prefer iteration idea rather than bool idea.
    • Cary: issue of intent, trying to capture intent of writer, not sure how you generalize that.
    • Li: passive container, wrong data is pipeline’s problem not exr’s problem.
    • Peter: EXR lumps together all 3 categories, standard and non-standard attribute and compression type. Ideally there would be 3 parts to the header.
    • Joseph: wrt integrity check, does it makes sense to define a metadata schema? Not necessarily part of the OpenEXR library.
  • No labels