Core/Universe Planning meeting 2021/12/22
Attendees
- Xinyu
- Fatih
- Ryohsuke
- Kenji
Discussion Topics:
Review Development Process using Core/Universe (Continued from ASWG)
Josh and I had a separate call to follow up on what we were discussing in ASWG.
Issue
- Josh was against the current plan to move forward with merging Autoware.Auto and Tier IV's AutowareArchitectureProposal(AAP) into Universe since they have different code quality.
- It would just going to be a disaster and cause confusion to the user.
Some of key discussion points
(These are made from my memory. I would like to have @JWhitleyWork to correct me if I am writing anything wrong)
- We both agreed that we should have a prototype phase in order to speed up the development speed, and universe would be a good place to do.
- Current Autoware.Auto has a lot of assumptions in implementation. If we merge Autoware.Auto into Core as is, it would take too much time and resources to refactor them through proper review that we expect for Core.
- But that doesn't necessarily mean that we have to have Tier IV's AAP to be in Universe
- The plan sounds like Tier IV's marketing to say that they have made contribution to Autoware.
- Ryohsuke: I don't deny that we don't have any of such aspect, but the main purpose is more about having more engineers to Tier IV to contribute Autoware development. Tier IV has it's own internal repository right now, but we spent time to make it Autoware.Auto compatible and trying to shift our development platform based on Core/Universe architecture. If we can have Tier IV engineers we would be able to relieve the lack of resources as well. We are also willing to modify our code according to the discussion made in https://gitlab.com/groups/autowarefoundation/autoware.auto/-/issues
- Josh: I understand the motivation from Tier IV's perspective. Considering the fact that Apex ceased from development of Autoware.Auto, it is likely that Tier IV would be taking the lead on development. If there are other main contributors in the community and opposing then we shouldn't be doing this, but if that's not the case I wouldn't be completely denying the plan.
Summary
- Josh and I got on the same page about the background and reasoning behind the plan.
- I think Josh is no longer fully against the plan( @JWhitleyWork please confirm )
- If we are moving forward with the plan:
- We still should have clear documentation about the Core/Universe strategy and be confirmed.
- We should make packages from Autoware.Auto is kept separate from other source codes with no proper review being done(e.g. by adding
auto_
prefix) - Once we have working core package, we should be removing duplicate package that came from AAP.
Go over documentation issues
- https://autoware-developer.slack.com/archives/C02L5379LBG/p1640173712023200
- https://github.com/autowarefoundation/autoware-documentation/issues/1
Action Item
Edited by mitsudome-r