2025-05-15 Meeting notes

Next Major Priority

• Node 22 for Ulmo

• Python 3.12 for Ulmo

• Focus on Deprecations to reduce the amount of code we have to maintain in the future

i

Retro: Studio rollout

@Kyle McCormick

Background

• The entirety of studio is nearly replaced with MFE frontends over the past ~2ish years

  • Largest recent change is the new Unit Editor in the Authoring MFE

  • Other than the unit editor, the rest of the MFEs had been released to edx.org for 6 months - 1 year

• The overall DEPR:https://github.com/openedx/edx-platform/issues/36275

  • Proposed Feb 19

  • Accepted Mar 14

  • Earliest breaking change (waffle flip) Apr 05

    • All MFEs would be enabled by default in the default Open edX configuration.

      • Because this was not true for the Open edX community prior to this in vannila openedx and a mixed (mostly MFE state) in Tutor

  • Earliest full removal Sep 14

• The PR in question:https://github.com/openedx/edx-platform/pull/36495

  • Merged Apr 24 - Previously planned for Apr 5th, switching to all MFEs by default in Studio

  • Announcement in the risky changes channel: https://axim-collaborative.slack.com/archives/C02QN50TYMD/p1745479221874829

  • This “released” several things for 2U, including Unit Page on edx.org, and several frontends on edge.edx.org

    • 2U had not added the opt-out waffle flags.

  • Opt out flags were added by Jeremy or some other 2U engineer.

Takeaways/Qs

• Waffle flags… to flip or not to flip (before rolling something out globally)?

  •  

• With opt-in flags you can easily add beta users to a feature, but the same is not true with the opt-out flag.

• Is there a better way to switch from opt-in to opt-out?

• The new DEPR process should make it more clear that operators need to take action.

• Doing this at the same time that Teak was getting cut where many other DEPRs were also being worked on and landed at the same time, making it difficult for operators running close to master to absorb all the changes.

• Another breaking change is coming in September and we don’t know how to represent that on the DEPR board.

• Why have two points of breaking change? On by default and removal of the old version.

  • Is there a way to easily change the default for everyone without making new waffle flags?

    • There is but it’s complicated because of course waffles and other things we have built on top of waffle.

• Is there a way to reduce the number of operator interventions?

  • We wanted to force an operator intervention to make sure they are paying attention to the upcoming major breaking change.

• Rather than making new waffles to transition, we should opt to make a data migration to rollout changes like this in the future.

  • Have we done this before?

    • way back when, yeshttps://github.com/openedx/edx-platform/blob/master/lms/djangoapps/grades/migrations/0018_add_waffle_flag_defaults.py

  • Why didn’t we do this more before?

    • It’s confusing to have long-lived waffle flags which default to True

• Unanswered

  • Should we have long-lived opt-out waffle flags ever?

    • https://github.com/openedx/edx-platform/blob/8c8a567dd0284bf820b0630b95c66399637dcef9/lms/djangoapps/user_tours/toggles.py#L15

Retro: Pre-Teak late-stage breaking changes to MFEs (FooterSlot)

@Adolfo Brandes

• History:

  • May 9, 2024: Original PR that added FooterSlot in frontend-component-footer, a breaking change only for forks of frontend-component-footer.

  • May 10, 2024: PR that undid the first one and instead implemented the footer slot in a non-breaking manner. Drawback: relied on a separate repository, frontend-slot-footer.

  • August 20, 2024: DEPR published for frontend-footer-slot, with the express intent of providing the footer slot in frontend-component-footer and removing frontend-slot-footer. Annoucement made in all relevant channels.

  • September 3, 2024: DEPR accepted.

  • April 17, 2025 (a few days prior to Teak cut from master):

    • PR merged that effectively implements the above DEPR as far as frontend-component-footer is concerned.

    • Announcement made on #cc-risky-changes that further PRs will be made (and merged shortly) to Tutor-supported MFEs to implement the change on consumers of frontend-component-footer

    • After Eastern Time Work day

    • Just before a 3 weekend in most of the world.

  • April 18, 2025:

    • Objection raised in the above Slack thread regarding a one-month window to handle breaking changes

    • A suggested fix for downstream forks is suggested by the author of the breaking change

  • April 25, 2025: Fix merged downstream to frontend-component-header-edx

Takeaways/Qs

• The new DEPR process should make it more clear that operators need to take action.

• The PR was merged on Thursday evening and required changes by the following Wednesday which left about 3 work days.

• There was more complexity in the change because the slot was being used in various MFEs that required updating to accept the new changes.

• Could people have done work before the PR merged or did the reaction need to be after the PR merged making it a lot more tightly coupled.