Checked exceptions (original) (raw)

Brian Goetz brian.goetz at oracle.com
Thu Oct 17 09:13:27 UTC 2019


I like Rusts model.

We're getting pretty far afield from the OP's question here ... but I think it's safe to say that the chances of Java embracing a radically different error model is nil.  As you point out:

If Java didn't have the legacy it has in such a huge API surface

and not only the huge API surface, but the huge surface of user experience and expectation.  But we do have those huge surfaces.

Further, while you might like Rust's model, nearly every model has strong attractors and detractors; all error models are uneasy trades between making it too easy to ignore errors and putting errors in your face.  It's not even the case that there is an "obviously better" model than checked+unchecked exceptions; there are plenty that any individual may think is obviously better, but that's not the same thing.  So we're not going to change horses here, especially given that all the models pretty much suck.  But we may file down the rough edges.

I think this blog post basically covers anything that might be said in this thread:

http://joeduffyblog.com/2016/02/07/the-error-model/



More information about the jdk-dev mailing list