[EMAIL PROTECTED] Wed, 30 Nov 2005 03:19:49 > >I know of several people who ran software mirroring >on Windows and they had >major problems with it along the lines that Greg >described. I also know some >people that never had problems in a similar setup >with OpenBSD. Prodded a >little more, they never had it crash so I guess >sometimes being stable can >actually hurt you ... :-) > You probably mean unstable, but it actually works both ways. My understanding is that the worst security problem is a false sense of security. Looks safe but isn't.
This does mean that the phenomena is not just theoretical. Tempting to say that the Windows code is buggy and the OBSD code is not buggy, but the results are completely consistent with the opposite case. The old-timers here keep urging an approach that is completely foreign to Windows. That approach, not the immediate code quality, explains the difference. (No offense to the OBSD developers, but watch what happens) Driving forces. Windows, anything which gives a slight performance advantage (impress the coder's boss) is a good thing, assuming it probably (probably is 50-percentile, about -1 nines) doesn't casue any obvious problems. OpenBSD has a different set of priorities. Maybe not so fast, but with any second-order complexities OpenBSD just might surprise you since effeciency is really more about never being too inefficient rather than sometimes being efficient. The OpenBSD implementation will as much as feasible try to not overly rely on things "being as the should". However, you listen closely to those who should know. Know what is assumedd and what the user needs to know to be survivable. Their instincts tend to be reliable even if they don't know just how or cannot express just how things go bump. I'd trust a 5-second off-the-wall reaction like that over an exhaustive and extensive 6-month detailed study. The Microsoft code wrecks disks, apparently consistently. It's probably safe to assume that Microsoft has put more (and arguably better by some standards) work into it. There are reasons that OpenBSD succeeds where Microsoft fails. I assure you the situation is not as simple as OBSD has good coders and Microsoft has bad coders. The situation is more like OBSD is NOT committed to making the mistakes that Microsoft is. This is essentially in the "attitude". The regulars know it. They've seen it. I've seen much the same kind of stuff in different contexts. No offence to the developers, they're good, much better than I, but they're not enough better to explain the differences. The approach is, and they're smart enough to have a good grasp of what they know, what they don't know, and that there is a difference. Where what how they do documentation. Without knowing entirely too much about everything, I'd trust their judgement. Everything I don't know matters. Not probably. For "probably" I'll use Windows. OpenBSD is for stuff that needs a positive number of nines in reliability. Making OpenBSD friendly like Windows. I think the best you can hope for is a very bad Windows. Moral of the story. (Certainly not just ccd) Listen to the old-timers on this list. Now why would I expect anyone to listen to me when obviously they never listen to said old-timers who DO know what they are talking about. End of rant. Short story.
