Does it though? Because when I read the actual attack / proof of concept it seemed to be that javaws was being run without validating the command line arg, which allows an attacker to pass their own -J param and path, and so boot with an additional jar.
Oracle seem to have ‘fixed’ the wrong thing - they’re trying to load the jar file and sandbox it after the fact with classloader hackery. But it strikes me that the proper root-cause fix would be to properly sanitise the command line before passing it to javaws so that rouge jar paths can’t be introduced in the first place.
That, combined with the no-beta quick roll out, has all the hallmarks of a non-technical manager throwing a hissy fit and forcing through such a colossally broken and ill-thought-out fix.
Edit: for reference, this is the original exploit doc I believe: http://seclists.org/fulldisclosure/2010/Apr/119