I believe this is a worthy discussion, perhaps also ultimately for the Platform mailing list.
I think we should try very hard to stick to mirroring certification results. Otherwise we enter a potential minefield of fairness, avoiding user confusion, potential devaluation of certification, weakening compatibility, etc. This is of
course especially acute as this is the official starter.
It would be good if others could weigh in on this. On Slack it was only myself, Scott, and Ondro that chimed in at all.
Let me raise the question from Slack:
https://eclipsefoundationhq.slack.com/archives/C047MCS83FT/p1678788937498749 to the list here:
Should the Eclipse Starter for Jakarta EE project use formal compatibility/certification/etc. to guard use of the UI and underlying archetypes, for a given combination of runtime+profile+version?
An alternative might be to allow someone working on a particular runtime to “vouch for” the usefulness of the function at a certain level if they work on the PR.
E.g. with Open Liberty close to EE 10 compliance, I wouldn’t expect Reza or one of the starter project committers to go out of their way to enable this, but if I personally (as
an Open Liberty committer) were to do the work to enable this then maybe my PR should be merged?
Some more thoughts:
Maybe there could be an asterisk (*) or warning/caveat about a runtime in such a not-yet-certified state?
If this is going to lead to debates about what should vs. shouldn’t be allowed though then perhaps it wouldn’t be worth it. A nice thing about keying off formal certification
is that we’ve already agreed to what counts as compatible.
I do see some potential for user confusion if uncertified runtimes are enabled via the starter.
But the counterargument is to consider the case that we’re very far along on the way to certify against EE 10, we’ve got a lot of useful function and this convenient starter… let
users take advantage of it. They don’t care that we’re disputing three TCK test methods at this point.
And assuming the runtime is in some kind of beta / early release then presumably there’s going to be a suitable “warning” implicit in the non-final version of the particular runtime.
Anyway, I could argue more with myself but let me send this out for discussion first.
Scott Kurz