Once again, there are technical reasons for this requirements and why it appeared in very early JavaEE times so don't think we should loose time discussing if it should be but we should more see how we handle the coming release and the fact current TCK are not compatible with this strong requirement.
I don't have any strong preference for the solution (1. release and ignore TCK until they are fixed ASAP after the release 2. have 1-2 people in task force mode to do the migration in a few days, 3. others) but it should be addressed - to be honest I didn't expect such a thread on that simple topic :(.
If any tck uses jakarta package it is a miss/bug in the javax->jakarta migration but platform didn't change so it must still be done.
If there is stated requirement in the Platform spec for applications to not add Jakarta package classes, the TCKs should respect that but if there is no such requirement, then the discussion should shift to focus on applications.
Any automated tool can do it if you want.
The impact if you don't do is that a vendor not passing TCK can still claim to be certified since TCK are not compatible with the spec, wouldn't be that great right?
The Platform TCKs validate the Platform spec requirements. As requested above, please provide section of Platform spec that requires that applications not create Jakarta namespace classes.
Le mar. 11 janv. 2022 à 01:03, Scott Marlow <smarlow@xxxxxxxxxx> a écrit :
On 1/10/22 5:04 PM, Dmitry Kornilov
wrote:
I’m lost in this conversation and surprised
that we are jumping to voting so fast.
The vote ballot as currently written is now for EE 11 only and
for new TCKs being added to EE 11, which means the answer to the
original question that initiated this discussion is that the Batch
TCK does not need any changes (nor does other TCKs being added to
EE 10).
I couldn’t find the driver for this
change.
The driver was a question about packaging for the Jakarta Batch
TCK.
I thought the discussion was jumping towards adding a Platform
requirement for applications to not define classes in the
"jakarta" namespace but we didn't seem to get there. IMO, the
time to make such an application restriction would of been for EE
9 (since making such a change for EE 10 or 11 would be a breaking
change now, even if it has low likely impact on most
applications).
I believe that less renaming is better for
the platform. Renaming and especially renaming in TCKs is a
big task. Who is going to do this? I am not ready to assign
people to it. Last time we changed names for Maven artifacts
we introduced the package split problems we still suffering
with. So, if you ask me -> NO RENAMING!
IMO, if application classes are allowed to define classes in the
jakarta namespace, than TCKs should also be allowed to do the
same. If we add a breaking Platform SPEC change to EE 10 or 11 to
not allow applications to define classes in the jakarta namespace,
TCKs should then be updated (new + existing) to follow that same
rule.
Scott
-- Dmitry
P.S.
Also, I don’t think that it’s up to the
platform group to make this decision. TCKs are a part of
specifications, so it must go through the spec committee to
get approved.
For
all new Jakarta TCKs, we need a
naming convention. All TCKs for
Jakarta EE 10, the requirement
is not to use jakarta.*. From
Jakarta EE 11, the naming
convention must be used.
The
new Jakarta EE TCKs
from Jakarta EE 11
must use the
following naming
convention. Pick
your choice below.