RE: JavaMemoryModel: Proposals on wait and interrupts

From: David Holmes (dholmes@dltech.com.au)
Date: Mon Jun 09 2003 - 19:18:06 EDT


Sylvia Else wrote:
> Well, yes, I can see that. However, I'm not pushing for a
> change. I'm pushing for the status-quo, in that I'm asking
> that the specification remain as it currently is. Any JVMs that
> don't implement that specification are faulty, and always have been.

But forcing the existing JVM's to adhere to the current specification
requires the very implementation changes that I was referring to.
Given a choice between changing the implementation in a non-trivial
way, to "fix" something that is of little practical concern, versus
relaxing the spec, it's not hard to see what choice is likely to be
made.

I'm not happy about this but I don't think we can avoid it.

I believe the VM must take responsibility for avoiding "lost
notifications", but that the relative order of interrupt vs. notify is
implementation dependent.

David Holmes

-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel



This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:46 EDT