Meeting notes 2020 October 26th #15
Conversation
meeting_notes/2020-10-26.md
Outdated
| - Olivier: HGBT and categorical features is still challenging; we don't need to block the PR and can handle the ease of use in future PRs. | ||
| - Check CircleCI checks: nice to fix all the failures for the release. | ||
| - documentation and nightly builds are no longer hosted on rackspace, but at anaconda.org ([#18652](https://github.com/scikit-learn/scikit-learn/pull/18652)) | ||
| - 1.0 Release: to be kept in mind when selecting or splitting PRs currently in 0.24 milestone. |
There was a problem hiding this comment.
I'm not sure I understand this point
There was a problem hiding this comment.
In my understanding, some complicated PRs are often split in small parts to make review and merging easier and faster.
This is ok when you move forward using a small steps approach: having 1.0 behind the corner, gives you a different perspective. Perhaps some problems could be solved if a PR is just abandoned and the problem completely reformulated: 1.0 is a good place to do that.
If this is not clear, I would be happy to remove... I'm not enough into scikit-learn API to give you examples.
There was a problem hiding this comment.
Thanks for the details.
We decided in the previous meetings that 1.0 would just be a very normal release without backward incompatibility (apart from the usual deprecations), so I'm not sure we can really change our incremental approach here (IIUC)
There was a problem hiding this comment.
We decided in the previous meetings that 1.0 would just be a very normal release without backward incompatibility
Well, definitely I'm having some hard time in following what decisions are taken during the meetings ;).
Anyway, I will remove this line, if this make people happier.
There was a problem hiding this comment.
you're not alone ;) that one was particularly confusing to me
some follow up here, for ref: scikit-learn/scikit-learn#14386 (comment)
No description provided.