3 ms·
In 2010, I recommended ballistic-missile-defense systems run their servers under both Linux and FreeBSD to mitigate the risk of kernel-level remote DoS 0-days [
by michaelngagnon 7y ago
In 2010, I recommended ballistic-missile-defense systems run their servers under both Linux and FreeBSD to mitigate the risk of kernel-level remote DoS 0-days [1]. Now it looks like N-version programming has failed yet again [2].
[1] http://michaelgagnon.me/file/isarcs.pdf http://michaelgagnon.me/file/isarcs.pdf
[2] https://en.wikipedia.org/wiki/N-version_programming https://en.wikipedia.org/wiki/N-version_programming
- sjackso 7y agoI read the situation as being worse for Linux (3 CVEs, including one remotely-inducable kernel panic, spanning many versions) than for FreeBSD (1 CVE that can be exploited to slow down a target system in the latest major version). But even granting that both affected OS flavors have serious DoS issues, a service that used a mix of OS flavors in its servers would be more resistant (not impervious) to attacks based on these TCP problems than a service using only one flavor. So do you really count this as a failure of N-version programming? (Honest question! This is not my area of expertise.)