IBM 1.4.0 JDK Incompatability

[dcartier]

Hi,

I recently decided to try the IBM 1.4.0 JDK on Linux again with the new versions of JBoss (3.0.7 & 3.2.0). I have tried using this in the past with older versions of JBoss but was stopped very early on in the boot process.

Now JBoss boots fine, yay!, but it seems IBM decided not to implement the "SHA1PRNG" algorithm. Doh!

After doing a bit of checking I have found that IBM has implemented an "IBMSecureRandom" (based on MD5 rather than SHA1).

I checked and JBoss appears to only use the SHA1PRNG in ~6 files, at least for 3.0.6, I do not have the 3.0.7 source handy.

So I was thinking, I could just hard code the algorithm selection to the IBM version, or perhaps we could include a bit more logic into the source and make the selection a compile time definition. Perhaps defaulting to the SUN version.

I also want to check to see if you specify the provider as 'SUN' if it will actuallly be able to find the SHA1PRNG even on the IBM JDK. Currently the code does not specify a provider.

Heck we could even have the code attempt the Sun version first and then try the IBM one if an exception is thrown.

I am reasonably new at Java, what does everyone else think?

Dennis

[dcartier]

Ok,

I have confirmed that changing it to 'IBMSecureRandom' fixes the generation of the clustered session id's. I have not checked the rest of the places it is used yet.

Adding 'SUN' as the provider was ineffective.

So either hardcoding (for myself) or making it a compile time option (for everyone) or having JBoss try the Sun implementation and falling through to the IBM (for everyone) seems to be the options.

If this is not the place to discuss this and it should be moved to a bug report, please inform me of this.

Dennis

[adrian.brock]

Make it a bug report.

Regards,
Adrian