Re: java.security.KeyPairGenerator oddity

[Casey Marshall]

>>>>> "Chris" == Chris Gray <address@hidden> writes:

Chris> Looking at KeyPairGenrator I see that one of the initialize()
Chris> methods is defined as follows:

Chris> public void initialize(int keysize, SecureRandom random) {
Chris> initialize(keysize, random);
Chris>   }

Chris> Either I'm missing something or this is a case of recursion:
Chris> see recursion. ;->

Chris> Sun's API docs are maddeningly vague as usual, but I _think_
Chris> that the default implementation should just throw an
Chris> UnsupportedOperationException.

Chris> BTW if anyone knows where to find a more useful description of
Chris> how this stuff is supposed to work I'd be grateful for a
Chris> pointer ...

KeyPairGenerator is one of the classes that were born broken when Sun
wrote them (MessageDigest suffers from a similar brain damage). I
think the point is that either (a) service providers write their own
`initialize' method when they extend the class, or (b) they implement
KeyPairGeneratorSpi and the getInstance method returns an "adapter"
class. See DummyKeyPairGenerator.

Throwing UnsupportedOperationException in this case sounds like a
better behavior. This is probably a case where it pays to emulate
Sun's behavior, but I don't think this will cause problems, since
service providers should extend KeyPairGeneratorSpi.

-- 
Casey Marshall || address@hidden