You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We have a few issues that came up from publishing recently, and since we hope to move other repositories to have published packages, this is a meta issue.
We published a major version of base-interfaces with a mistake, which forced TWO major version releases in the span of 30 minutes. That' no good.
We recently learned that we were missing a build step in our publishing processes, which meant that version 1.1.0 of base-constants did not actually include the new features in the published package.
We are going to make mistakes, so really this is a problem with our process.
First: I would like us to have processes around publishing release candidates / beta / alpha versions for any breaking changes.
Second: since we're in a period of rapid development, and we accidentally skipped past the 0.*.* versioning that SemVer allocated exactly for rapid development, I would urge us to put ANY major change into a beta / pre-release with the intent that we'll "live" on the pre-release until our APIs are locked in / we get past the fluid stage.
Third: We should document our release processes, this can / should be done in the meta repo; I'm comfortable with either a wiki or a committed RELEASE.md file. We should make an issue to capture this request.
Fourth: We should look into improving our automation, especially in monorepos. We ditched lerna (which I think is good) but now publishing is a manual and therefore error prone process. I'm uncomfortable with it all!
Task
Description
We have a few issues that came up from publishing recently, and since we hope to move other repositories to have published packages, this is a meta issue.
We published a major version of
base-interfaces
with a mistake, which forced TWO major version releases in the span of 30 minutes. That' no good.We recently learned that we were missing a
build
step in our publishing processes, which meant that version1.1.0
ofbase-constants
did not actually include the new features in the published package.We are going to make mistakes, so really this is a problem with our process.
First: I would like us to have processes around publishing release candidates / beta / alpha versions for any breaking changes.
Second: since we're in a period of rapid development, and we accidentally skipped past the
0.*.*
versioning that SemVer allocated exactly for rapid development, I would urge us to put ANY major change into a beta / pre-release with the intent that we'll "live" on the pre-release until our APIs are locked in / we get past the fluid stage.Third: We should document our release processes, this can / should be done in the meta repo; I'm comfortable with either a wiki or a committed RELEASE.md file. We should make an issue to capture this request.
Fourth: We should look into improving our automation, especially in monorepos. We ditched lerna (which I think is good) but now publishing is a manual and therefore error prone process. I'm uncomfortable with it all!
Relevant Resources / Research
Related Issues
None.
The text was updated successfully, but these errors were encountered: