torriem at gmail.com
Wed Nov 18 09:46:04 MST 2009
Steven Alligood wrote:
> My guess is that the windows servers listed at the above url are
> actually clusters, that allow upgrades, reboots and testing of one at a
> time offline. Not true uptime at all.
This reminds me of an article I read recently about software maintenance:
It still galls me to have to reboot my Mac every month or so because of
critical software updates. Maybe HFS+ can't replace in-use files like a
normal inode filesystem can. Certainly that's the excuse Windows uses.
In both Apple and Microsoft's case, however, it's pure laziness on
their part, plain and simple. If you can save a programmer 40 hours of
work by making customers bear the cost in time and money thousands of
times over instead, they do it.
But even on linux, a kernel update requires a reboot. Often the kernel
update is critical because of a local exploit that it fixes. Why do we
have to reboot just to patch a kernel? Sure it sounds complicated to
patch a running kernel, but if I recall there were systems in the 70s
that could do this. There must be mechanisms that could be used to
facilitate this in modern Linux kernels. But like the Microsoft
programmers, Linux programmers (aren't we all) are inherently lazy and
shift the costs in a similar way.
Seems to me if we focused on software maintenance (which really means
more than bug-fixing; think evolution), not only would uptimes be
ridiculously long, but software in general would be more reliable. But
that'd go against human nature.
More information about the PLUG