> > Enabling a world-wide auto-update feature does indeed seem much of a > > security risk to me. > > > More of a risk than up2date for RedHat or emerge -u system for Gentoo? Or > cvsup for *BSD? Yes. These systems are voluntary. The structure of UNIX systems, and updates makes it much easier to test and update and less likely to kill a system even if it is flawed. Updating user space applications in Red Hat, other then SSH causes me essentially no nervousness. If Apache bombs out I can trivially roll it back to an older version, or if it's totally screwed up remove, and replace (and not lose my config files either since most RPM packages are designed to protect old/original config files). I cannot selectively install say 90% of Service Pack 4 in Windows 2000, it's pretty much all or nothing. In Red Hat (and Linux/BSD in general) there are no roll up security fixes (exceptions being appliance vendors, but those are much more tightly bound environments and less likely to suffer update related problems). To put it bluntly: Compare the number of times MS has had to re-release patches/updates or additional fixes because it kills/breaks something vs.s. the number of times Red Hat or SuSE does. P.S. even if MS does solve most of these issues they still have painted themselves into a corner due to their rather insane file locking which requires a reboot in virtually every circumstance you want to replace important files. > Paul Schmehl (pauls@utdallas.edu) Kurt Seifried, kurt@seifried.org A15B BEE5 B391 B9AD B0EF AEB0 AD63 0B4E AD56 E574 http://seifried.org/security/