robustness vs. conservative GC

[Boehm, Hans]

> -----Original Message-----
> From: Tom Lord [ mailto:lord@regexps.com ]
> 
> After reading the omitted parts of your reply, I stand by original
> statement, except that it would have been more accurate to say
> "unpreventable" instead of "uncontrollable", since the alluded to
> failures can be deliberately caused -- a form of control which
> undermines the value of conservative collectors in numerous
> applications.
> 
You're talking about a denial of service attack on say, a Java
implementation, by causing it to retain extra memory?  I don't think that's
fundamentally different from retaining lots of memory with real pointers.

Or are you talking about a GC-unsafe Java compiler used with a conservative
collector?  I'd argue that's fundamentally no different from another (very
rarely triggered) compiler bug.

Hans