While fixing backwards compatibility for DMN files with camunda namespaces, I realized that the CMMN parser is not backwards compatible, either. CMMN was already unsupported in the later Camunda versions, and right now there are / were no plans to pick up the support for it again. I know some people still use the feature or have legacy case definitions deployed.
My proposal for now is to only allow CMMN definitions with the “old” Camunda Namespaces and not make the CMMN parser read XML files with a new “Operaton” namespace. On the one hand this signals that we do not pick up work on CMMN again, on the other hand it won’t break migrations of applications still using the unsupported CMMN feature.
I think we can either drop support for CMMN now, or it will never be dropped. Meaning someone has to maintain it. We - as a user- don‘t even have CMMN models we could QA test the software with.
I think it’s an interesting feature and has a similar justification in the engine as DMN. For now, I wouldn’t put any direct effort into it, but I wouldn’t want to remove it either.
Another question would be: will CMMN be available in the renewed cockpit Web App? If we want to sunset the old one and this is a supported feature, there must be a replacement
From a product management perspective, it’s clear that we need to have some discussions about how to evolve (or not evolve) the case management capabilities in the long term.
We’re fortunate to have experts in our community who can provide valuable insights. For now, keeping things as they are might be the best solution until we’ve had a chance to explore the use cases and understand them better.
I have actually made a presentation on how to solve the need for CMMN witrh BPMN and events.
Unfortunately its in Finnish.
So I would say, just dump the CMMN or at most, allow exisitng models to work, but do not put effort into developing it.
Indeed, CMMN offers a few capabilities that BPMN does not. However, as developers of a case management product, we only use BPMN. It’s much easier to model and more understandable. There are some limitations, but in my opinion, they are not significant enough to justify using CMMN. I would drop it.