Re: Plan regarding RC4

At 20:44 20/05/2003, Sascha Schumann wrote:

On Tue, 20 May 2003, Zeev Suraski wrote: At 19:24 20/05/2003, Sascha Schumann wrote:

    The changes are not as massive as Zeev wants you to believe.

How come you introduced 3 crash bugs with this extremely stable patch then?

    Your way of counting is misleading.  The changes uncovered
    bugs in existing code (lack of proper initialization).  All
    these have been fixed.

That's wrong. The semantics of file handles was that they do NOT require initialization. You changed that, and introduced these bugs. You simply misunderstood the code and didn't expect it to behave the way it did. The end user couldn't and shouldn't care less. And even that only relates into one of the crash bugs.

    The require_once/include_once problem was indeed caused by
    "late" changes.  I initially missed this piece of code
    altogether, so that it was not as properly tested as the rest
    of the affected infrastructure.  This has been rectified as
    of today, and thus I cannot share your risk assessment.

The fact that you're quick to fix the bugs once they're found is not related in any way to the risks involved with this patch. The introduction of three crash bugs is just a good illustration of what *can* happen. It would have been poor risk management to introduce this patch even if we didn't have three real world examples of what it can cause. Since these three cases did exist, though, explaining why it's bad risk management is a much simpler task. Zeev