IMHO we should use the lowest supported version of Scala 3 to not force user to upgrade to newer versions.
Scala 3 continues to deprecate old syntax. Some of it will produce warnings in Scala 3.2 and may be removed in 3.3 or later. For example https://dotty.epfl.ch/docs/reference/dropped-features/package-objects.html or https://dotty.epfl.ch/docs/reference/changed-features/imports.html One precondition to upgrade to newer versions of Scala 3 would be dropping support for Scala 2.12. Scala 2.13 at least has support for some of the Scala 3 Syntax with compiler flags to cross compile. On Fri, 24 Mar 2023 at 10:26, Matthew Benedict de Detrich <matthew.dedetr...@aiven.io.invalid> wrote: > So some discussions on github are popping up regarding which Scala 3 > version we should pick so I think it's time to discuss this formally on the > mailing list. > > As a precursor one thing people need to understand is that the Scala 3 > release cycle has changed, quoting from > https://github.com/apache/incubator-pekko/issues/6#issuecomment-1302701657 > > > Scala 2 used epoch.major.minor version convention. Scala 3 has > major.minor.patch. > > > So there is no 3.0/3.1/3.2/etc cross-compilation - the assumption is > that: > > > * you can compile against the same minor version with backward- and > forward-compatibility: 3.1.3 dependency against 3.1.0 code, 3.0.0 > dependency against 3.0.1 code, etc > > * within the same major version you always have backward-compatibility: > 3.1.3 dependency can be used in 3.1.3 project, but also 3.2.0 project and > in future against 3.3.0 project > > This means that if we pick a Scala version, we are essentially forcing the > potential Scala 3 users of Pekko to bump their Scala 3 version to the minor > that we decide on. On surface value this means that we should pick the > lowest Scala 3 minor version that we can support however there is the fact > that Scala 3.3 is going to come out soon which will be an LTS release. The > LTS release means that if any bugs are found after Scala 3.3 for a period > of 2 years, they will be backported to Scala 3.3. The big advantage this > brings us, is that it allows us to freely bump Scala 3.3 without breaking > our users if any potential bugs are found in the future. If we decide to > stick with Scala 3.2 or 3.1 and some bug is found in Scala 3 later on that > affects us, we will have to update to a version of Scala 3 that will break > binary compatibility. This facet is of even more importance when > considering our 1.0.x release branches, which are designed to never break > binary/backwards compatibility, i.e. if we do 1.0.x releases with Scala > 3.1/3.2 and some critical bug/CVE comes out later we could potentially be > forced to update the minor version which would break this binary/backwards > compatibility, this wouldn't be the case with Scala 3.3 (for a certain > period of time). > > Of course the counter argument to using Scala 3.3 is that it would force > all potential Pekko users (and the transitive set of Scala 3 libraries for > that Pekko user) to also use/support Scala 3.3. Unfortunately its not > possible to get download stats from Sonatype for artifacts you don't > maintain, but I wouldn't say its a controversial statement that the amount > of people that use Akka long with Scala 3 would be a tiny minority (this is > also regarding other factors, i.e. the typical demographic of Akka users). > Ontop of this we need to take into account the delay of current Akka users > migrating to Pekko, in other words by the time people migrate to using > Pekko the fact that its using Scala 3.3 LTS would likely be a non concern > at that point in time. > > And finally another thing to note is that even in the worst case scenario, > nothing is stopping users from using Scala 2 artifacts from within Scala 3 > (this is perfectly supported and has been for a while). Afaik the current > Scala 3 version of Akka/Pekko is not using any unique/bespoke features of > Scala 3, if true this would mean from a Scala 3 user perspective there > really isn't going to be > a difference in using a Scala 2 artifact vs Scala 3 artifact. > > For these reasons my recommendation would be, assuming that the full > release of Scala 3.3 LTS is ready by the time we decide to make a release > that we should try and target that. For details on the current release > schedule for Scala 3.3 LTS you can read > https://contributors.scala-lang.org/t/3-3-0-release-thread/6079/3. > > -- > > Matthew de Detrich > > *Aiven Deutschland GmbH* > > Immanuelkirchstraße 26, 10405 Berlin > > Amtsgericht Charlottenburg, HRB 209739 B > > Geschäftsführer: Oskari Saarenmaa & Hannu Valtonen > > *m:* +491603708037 > > *w:* aiven.io *e:* matthew.dedetr...@aiven.io >
