enable onError without enabling alpha API

[pritidesai]

Pipeline 0.27 was release in August which introduced an alpha feature to specify onError in a step. This feature was implemented in the PR #4106. To enable this feature, the Tekton Pipeline deployment has to set enable-api-fields to alpha.

In our deployment, we have many use cases which require the functionality onError provides but cannot enable alpha APIs.

What do folks think about introducing an additional feature flag to just enable onError functionality?

An alternative is to promote it to stable/beta API. Based on our guidelines, promotion can be done after a few releases(3 to 4). Can we file an exception for this feature?

/cc @skaegi @bobcatfish @vdemeester @afrittoli @sbwsg @jerop

[jerop]

I prefer adding an additional feature flag to promoting it to beta faster so that we have room to easily make any necessary changes based on user feedback

More broadly, I wonder if we should have guidelines for these circumstances considering the cons of individual feature flags as discussed in Feature Gates TEP, such as:

• do we want to consider adding individual feature flags on a case by case basis?
• do we want to add tooling to show which individual features have been enabled?
• do we want to set expectations for testing coverage of the individual feature with other related features?

[jerop]

couldn't find a central documentation for all alpha features so added it here #4246 - please take a look :)