Call for Discussion: MSI installer for Windows

[Alex Kashchenko]

Hi Alexey,

On 6/27/22, Alexey Semenyuk <alexey.semenyuk at oracle.com> wrote:
> Hi Alex,
>
> OpenJDK already provides a packaging tool for Java applications and Java
> runtimes.
> We don't believe OpenJDK needs another one targeting a narrower scope.
>
> Rather than contributing an MSI installer to OpenJDK which would
> duplicate a lot of the functionality already in jpackage, we'd encourage
> you to contribute to jpackage enhancements allowing it to satisfy
> broader range of requirements for Java runtime installers and should
> allow any OpenJDK vendor to leverage it to create whatever exact
> installer they may want.

Thanks for your comments!

While I may disagree about the "duplicate functionality", I would like
to note that jpackage enhancements (outlined in [1], but not limited
with the features listed there) were intended to be contributed as a
prerequisite to possible MSI installer make target. I believe, that
such enhancements to jpackage (in runtime installers area) can be
beneficial, regardless of whether the make target is added or not.

I also think, that some kind of "vanilla upstream" installer (produced
by jpackage from "make images" output, with a concrete fixed set of
input flags/resources) may be useful as a baseline for vendor-specific
installers. For example, we at Red Hat would prefer to maintain the
installer as close as possible to (supposed) vanilla upstream one.

I intend to continue with jpackage contributions, with a goal to be
able to create installers with the set of features that match existing
LTS Red Hat installers. Hopefully the idea of a "vanilla upstream"
installer JEP may be reconsidered in future.


[1] https://github.com/akashche/jdk/blob/installer-msi-jpackage/make/data/installermsi/doc/JPACKAGE.md#jpackage-additional-features

-- 
-Alex