[concurrency-interest] Finalization changes to DelegatedExecutorService

Brian Goetz brian at quiotix.com
Thu Mar 23 13:17:50 EST 2006


> (The broader moral is that even when you purposely do not specify some
> behaviors, people come to rely on them, so it becomes harder and harder to
> change anything at all as a set of APIs become established.)

And also that no matter how hard we try, platform and performance issues 
influence the design of APIs.  We restricted ourselves to the choice of 
having newFixedThreadPool return ThreadPoolExecutor, or ExecutorService; 
we chose the interface because of the two, it was the sensible thing to 
do for a number of reasons.  But what we really wanted was some sort of 
ConfigurableThreadPool interface, which we were reluctant to define 
solely for this purpose because of concerns over class count, which we 
were concerned about because of the need to make this stuff useful on 
constrained platforms like J2ME.




More information about the Concurrency-interest mailing list