Hi Charles and Jim.
Thanks for starting this very refreshing topic. And thanks above all to you @chuckyboilo, for thinking about posible archival strategies.
Operational points first:
- I think that archiving may selectively fulfill a rather wide range of objectives and, depending on each goals, the archive will be different (different assets to archive, with different encodings, file formats, naming conventions, etc.). As Jim said, we’re still far away from a fully-tagged, metadata-driven “übermaster” that completely describes itself, so we’re still at the point where tactic is heavly dependend on specific archival objectives; different facilities, or productions may thus require different types of archival copies.
I’m not implying any project should be archived in a different way: but there may be a ver few selected standards tailored to the main families of objectives (read below two items).
I see IMF being a technology going in the right way though, especially due to its XML-based expandability and sub-standards declined as “Applications”. - Archival masters for either legal/compliance, re-broadcasts, or film-heritage purposes, for example, may need neither ungraded, nor textless, nor “handle-frame” clips; their presence in the package might even be detrimental when primary goal is an asset faithful to a specific edition of the content (e.g. the original domestic theatrical release).
- Archival copies for preservation, localization or technological re-release, instead, may very well benefit from such “open-structure” packages allowing things such as re-editing and re-grading.
Technical comments now:
- IMF prescribe the use of MXF container format for storing the essences; each application may demand metadata and specific encodings for essence wrapped within the MXFs of the package. JPEG2000 is used for some Applications – not all.
- Using a standard colorimetry for every master/archival copy –as is the case of ACES2065 (AP0 primaries) for ACES– is in my opinion better than preserving the original camera-native files, because of the widest gamut and because there won’t be in the future any need for software that can interpret raw file formats with possibly-obsolesced, propietary and/or undisclosed features such as Bayern pattern, camera metadata and camera-native colorimetry, etc…
- Standard ST2067-50 IMF, Application #5: ACES, for example, prescribes to encode video as frame-wrapped, with each frame encoded as OpenEXR. The details for either the underlying EXR and the MXF structure on top for App #5, are referenced in ACES family of standards ST2065-4 and ST2065-5, respectively.
- In IMF App#5 reference frames can be added, as individual PNGs or TIFFs, encoded in (possibly different) output-referred color-spaces. They relate to reference display(s) used during mastering of specific versions. They encode the codevalues of those frames after applying the relevant Output Transform(s), to serve as visual and numeric reference even years in the future. For this reason using at least one open-spec colorimetry (that can exist or be replicated even when no such display technologies exist any longer) is a good practive, in my opinion.
By the way: It’s not yet confirmed, but I should be giving a talk at this year’s Cinema Ritrovato festival (June 2018) about archival using either ACES and IMF from a preservationist’s perspective.