On Thursday, May 19, 2016 at 3:13:44 AM UTC+5:30, bream wrote: > On Wednesday, May 18, 2016 at 6:47:42 PM UTC+1, Chris Kaynor wrote: > > On Wed, May 18, 2016 at 10:15 AM, Steven D'Aprano wrote: > > > > > I don't really understand why the system can't track the current top of > > > the > > > stack and bottom of the heap, and if they're going to collide, halt the > > > process. That would still be kinda awful, in a sudden "your application > > > just died" kind of way, but it would be better than "your computer is now > > > owned by some hacker in Hong Kong, who is now renting it by the hour to > > > some spammer in Texas". > > > > > > > Most modern OSs will track it, and kill the app (hence the exception/crash > > that occurs), rather than allow access outside the memory. > > Chris > > Oh so funny, I just had to roar with laughter, unless you (plural) classify > VMS as a "modern OS". Then of course some youngsters think the OS is simply > software, whereas us old gits know that a combination of hardware and > software makes for something that is rather more solid.
Hi Mark -- howdy! Much truth in your comment. Dijkstra pointed out (in mid-70s!) the essential tradeoff between hardware and software. He did it for numeric oveflow but it applies equally for stack overflow. In short software costs are variable costs: An instruction that is executed a million times costs million-fold one that is executed only once. For hardware its a fixed cost: Once the investment in silicon is done, once or a billion times is the same. Therefore -- and this is Dijkstra's key insight -- if a test is very skew, ie if the if-part is 99.9% likely to be taken and the else part only 0.1% the programmer will be increasingly tempted to drop the test and hope/pray that the 0.1% does not happen. With a hardware interrupt this temptation would be obviated. Both the mess in catching numeric overflow as well as stackoverflow looks like its C's fault. I consider it as the fault of currently fashionable stock hardware -- https://mail.python.org/mailman/listinfo/python-list
