Given the increasing tempo of build failures to to npm, the challenges with shrinkwrap and the inability to use mirrors, it may be time to evaluate the use of yarn as an alternative.

Approach

• General things to validate
  • Ensure "run scripts" still work with yarn
  • maven-frontend-plugin supports yarn natively, so we just need to switch this over.
  • Do local installs work nicely? (for rapid development). Can we obviate or at least better integrate w/ slink so rapid dev is more seamless? Check out "yarn link"
• JDL Next
  • Need to determine if/how a Yarn-published module is able to freeze its dependencies when pulled in via npm (since there is no shrink, probably not, but maybe there's some magic)
  • Publish some beta JDL's and then pull them in via a blueocean-plugin PR to ensure ATH passes and things work, etc
• blueocean-core-js
  • Do the same as above
  • At this point perhaps we add JDL as a peerDep and avoid the need to republish core-js each time new JDL is published (large pain point)
• blueocean-web, dashboard and personalization
  • Port these over to yarn as well
• Test with Artifactory