Experience with rules engines

We are considering adding a Rules Engine to our architecture.

From so开发者_如何学Pythonme reference we hear positive stories.

From others we hear that it is more work than it is worth, and that some companies have given up.

Do anyone have a sucess or a horror story that they could share?

I've worked with rules engines from different vendors (TIBCO, Corticon), and if you accept their limitations, they can be useful. They allow business to change processes easily, directly, without IT, which is good and bad at the same time.

The first story is a landline telecom operator setting up order decomposition with decision tables. None of the commercial rules engines provided the necessary functionality, so they built their own, using DB tables, triggers and stored procedures. Works all right, but only a couple of people understands how - although the original idea was to use decision tables, because they are simple...

The second story is an insurance company, building an online questionnaire for reporting sick leaves. This is the typical application for rules engines, still did not succeed. Business was happy they can change the rules, there was even an approval workflow for the changes. The rules engine then generated some Java classes that the decisions server could import without restarting. IT Operations rejected the solution, because a new Class file is a code change, that needs to go through an official acceptance test cycle (several months + $$$). Insurance company ended up using simple DB tables and updates as SQL scripts.