I agree to remove Inifinispan, although I have invested a substantial amount of work to keep it working. But we have better alternatives for embedded, both relational (embedded postgresql, h2) and non relational (rocksbd) Also, it does not have the same level of query support than Mongo and Postgress (it is not possible to query variables or metadata) Besides, I also find questionable the original decision to use it as default data store, given the fact that the optimal use for Inifnispan is as persistent distributed cache (so, in a way, it was like using Kafka as datastore, possible, but not its original intent) I think is good to deprecate it and focus on keeping JPA and mongoDB implementation at the same level of functionality (currently Mongo is slightly behind) Thats said, I think that, rather than removing, we can exclude it (just remove the reference from the parent POM and keep the code)
On Tue, Apr 1, 2025 at 2:02 PM Alex Porcelli <a...@porcelli.me> wrote: > I think those interfaces are already in place - at some extent. We have a > few different db implementations already: mongo, Postgres, rocksdb; so I’d > suspect that it should be enough for now. > > I also would argue that H2 is also a reasonable solution for in-memory with > low cost and is now actively used in jBPM when running in dev mode. > > - > Alex > > On Tue, Apr 1, 2025 at 7:54 AM Gabriele Cardosi < > gabriele.card...@gmail.com> > wrote: > > > HI Alex, > > I agree with your concern. > > I do not know exactly where and why infinispan is used, but I also hit > > issues with it during some upgrades. > > From my POV the issue is that > > 1. On one side, we want to provide in-memory support for our > applications, > > which is a good feature > > 2. on the other side, we implemented specific solutions, with > hard-binding > > to that specific implementation (infinispan), that lead to the issue you > > mentioned > > > > Being everything open source, maybe a good solution would be to design > and > > provide API/interfaces for any "in-memory" database, and then whoever > wants > > to create a specific implementation, outside our apache-kie project, > would > > still be able to create it. > > > > Wdyt ? > > > > Il giorno mar 1 apr 2025 alle ore 13:23 Alex Porcelli < > porce...@apache.org > > > > > ha scritto: > > > > > As part of the ongoing effort to upgrade to the latest Quarkus LTS > > version, > > > we’re currently dealing with the impact of aligning with Infinispan 15, > > > which introduces a new set of changes and potentially api compatibility > > > issues. > > > > > > Given this context, I’d like to revisit the question: Do we really want > > to > > > continue supporting Infinispan? > > > > > > This has been discussed in the past, and there was some resistance to > > > removing it. However, maintaining support goes beyond the existing > > > implementation—it requires us to stay on top of future upgrades, adapt > to > > > API changes, and deal with potential security vulnerabilities stemming > > from > > > Infinispan itself and its transitive dependencies. > > > > > > I believe it’s worth re-evaluating its value and whether it’s aligned > > with > > > the future direction of the project. > > > > > > Looking forward to your thoughts. > > > > > > - > > > > > > Alex > > > > > >
