General updates | JW – Something about our docker subscription lapsed, so I’ll need to move the QC worker docker images and conda env yamls to somewhere else - likely ghcr (github container registry?). We currently use dockerhub. Happy to do a screenshare on that if you’d like to see how it works. Next Monday is labor day, then the two mondays after that I’m in europe. Could we cancel and then schedule a meeting if needed?
BW – Benchmarking runners/actions JW – We can discuss major design stuff/decisions at the internal benchmarking meeting, but what are some details that we can work out now? BW – Migration to OFF GH org and final naming? BW – How to use OE license in secrets? Call for user stories for Weds? Weds meeting items: Provenance requirements? Set a timeline for versioning/stability? (eg “this is unstable for the first 3 months, don’t expect to be able to reproduce results from that time…”) Should zenodo records document how to use themselves? (eg “create an env from this yaml gere by running …, then load the sqlite file here…”) Format of uploads? (one big zip file? A bunch of loose files?) BW will check whether it’s possible to fetch a single file from a larger zenodo record) Manual review before zenodo publication? (Automation can create a draft and require human review before cutting) Create an index mapping submissions/folders/PRs to zenodo record? JW – Could have a table in the GH repo where each submission is linked to its PR, folder, and zenodo. And links from zenodo back to the GH repo. BW – Maybe jinja could update this table. Would be tough to have it in the main README without rewriting it each time. So maybe a big index in a different folder/file.
Validation machinery - make sure input is valid-ish How to trigger dataset computation and upload? GH comments (current version) PR merging (would require a rework) Relabeling (what qca-dataset-submission does)
Processes for… Deduplication logic/what to do if files would be overwritten in GH If we use Zenodo, what should actually be committed to GH in a PR (or do we just use PRs for decision/review process?) What’s the lifecycle? User submits PR PR gets approved by human CI 1 is kicked off using some human trigger CI 1 runs YAMMBS job CI 1 prepares zenodo draft CI 1 updates index of jobs CI 1 finishes, comments on PR to tell user next step The draft is somehow published
BW will think through this workflow some more and put a diagram together for weds BW will prepare a test run and zenodo draft for wednesday
|