Java memory model

[Guy Steele - Sun Microsystems Labs <address@hidden>]

Maybe it will help to douse the flames (with H2O or C8H18?)
if I remark that, in my opinion, the memory model of the
Java programming language is broken.  Consider it so remarked.

I am very proud that Bill Joy and I tackled the problem of
specifying a memory model for an explicitly multithreaded
programming language instead of just ignoring it and hoping
for the best.  I think we broke new ground here in the art of
programming language specification.

I am embarrassed that the model we set forward has certain
obvious technical and practical flaws, and certain other very
subtle flaws.

I am grateful that Bill Pugh has worked so hard to clean up the
mess and propose an alternative memory model that looks very
promising.

I am hopeful that, once a revised memory model is adopted,
implementors will make an effort to ensure that their
implementations obey the memory model instead of just ignoring
it and hoping for the best.

--Guy