Skip to content

Conversation

@hogepodge
Copy link
Contributor

No description provided.

@hsaputra
Copy link

Are we not sending the copy of topics from discuss.tvm.apache.org to dev@ mailing list anymore?

I could not find the email copy for the TVM discuss that linked with this PR: https://discuss.tvm.apache.org/t/rfc-update-rfc-process/9033

@hogepodge
Copy link
Contributor Author

It looks like the forum is being archived. I don't know why that particular topic was missed.

@hsaputra
Copy link

Ah, I saw your latest comment from the TVM archiver to dev@ list. Maybe I accidentally delete the previous emails. I think we are still good. Thanks, Chris

@tqchen
Copy link
Member

tqchen commented Mar 31, 2021

cc @apache/tvm-committers

@tqchen tqchen added the status: need review RFC needs review label Mar 31, 2021
@tqchen tqchen changed the title RFC Process Update [RFC] RFC Process Update Mar 31, 2021
Copy link
Contributor

@areusch areusch left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

doesn't have to be part of this change, but very supportive of adding a linter to this e.g. via GH actions or script + docker container. it would also be great to think about a policy as to how to link to things. For example, when linking to other RFCs, should links go to this repo or to the discuss forum?

Related is my comment on identifying RFCs. If RFCs can be easily identified by e.g. a number or name, then linking to them in this repo is also trivial.

@hogepodge
Copy link
Contributor Author

@areusch @tqchen @jroesch ready for final review

README.md Outdated
updated to reflect the change. In the instance where the change is a
significant addition rather than a simple modification, a new RFC should be
posted.
- **Postponement**: An RFC may be postponed either
Copy link
Member

@tqchen tqchen Apr 22, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems to be a departure from the rust RFC model https://github.com/rust-lang/rfcs. I would recommend tomake the process more consistent with the rust RFC process(given it is already adopted as a good practice). In particular, a RFC pull request itself can be postponed, but the implementation of RFC should follow a separate process. See https://github.com/rust-lang/rfcs#implementing-an-rfc

Maintaining the state of the RFC implemention inside the rfc repo can also be a maintaince overhead. The status implementation should be tracked in a tracking issue.

Normally some RFCs are implemented immediately and some are done through out a longer process, however, once an RFC is accepted

@jroesch
Copy link
Member

jroesch commented Apr 22, 2021

I think we just need to resolve @tqchen's comment since I think the request was brought up by @areusch and we can merge.

@hogepodge
Copy link
Contributor Author

@tqchen @jroesch I updated the process to more closely match the Rust process. A few comments on it:

  • I still kept it more lightweight than the Rust process. As a smaller community, I want to strike a balance between formalism and ease of application.
  • One major difference is in postponement. The Rust community appears to postpone RFCs before they are merged, adding a tag. This is in contract to TVM postponement, which is meant to capture features that may have been accepted but aren't actively being developed. If this is a point of confusion, I would suggest we change the name to something else. I'm not convinced that the Rust model of postponement is necessary for a project of this size and scope.

@tqchen
Copy link
Member

tqchen commented Apr 23, 2021

One concern is the need to move RFC text around depending on its implementation status. In practice it would add additional maintainace overheads and also loses the history of the RFC moditication. Per Rust RFC process, merging an RFC means that the community has get consensus on the design and desired way of improvements. Ideally the concensus of the design should be independent from the implementation status. Ideally the path to the text should not change after the RFC has been merged.

The actual state of implementation can be tracked in somewhere else(e.g. a tracking issue) independent from the RFC repo(which reflects the consensus on the design). Alternatively, an RFC itself could contain a short status section that shows (proposed, implemented). I would be cautious marking an accepted RFC as postponed unless the design decision is superseded and no longer holds, in which case we might mark the rfc as abandoned.

An implementation of RFC can happen soon, after a few month depending on the progress, they can also happen asynchrously given that apache-style development is quite asynchrous.

@hogepodge
Copy link
Contributor Author

Ok, that clears things up for me. Thanks. @tqchen I've updated the PR to move more of the mechanics of handling the RFC into the tracking issue, and deleted the requirements for moving the RFC around.

@tqchen
Copy link
Member

tqchen commented Apr 24, 2021

Thanks @hogepodge . It looks good to me now. Now that we have close to finalization of the text, would be good to invite more community members to take a look by creating a new discuss post

@hogepodge
Copy link
Contributor Author

Doesn't look like there's much conversation around this. I'm ready to move forward on it when you think it's appropriate.

https://discuss.tvm.apache.org/t/rfc-preparing-to-launch-new-rfc-process/9838

@tqchen
Copy link
Member

tqchen commented Apr 29, 2021

cc @apache/tvm-committers final call for feedbacks for 36 hours before we merge and start the formal voting process

@tqchen tqchen merged commit 18937f4 into apache:main May 6, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

status: need review RFC needs review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants