At one stage there was some uncertainty as to whether he was a bug or a feature. Apparently because FreeBSD used to have a page about him, it was decided he was a feature.
After re-reading both accounts, it's clear to me that Brett misunderstood Poul-Henning's argument. Poul-Henning said to "just build the damn bikeshed, I don't care what color it is," whereas Brett said, "Poul-Henning's assertion that all such ideas should be dismissed as "bikeshedding" reflects this dismissive attitude"
I read Poul-Henning's argument, which did not say that ideas should be dismissed, but rather 'just built', as attempting to contribute to a constructive spirit, and Glass's argument as defending the broken management process.
It depends on how sleep(3) is implemented really. Usually sleep means approximately one second resolution. If you start adding fractional second sleeps, say 1.5ms is required, you start moving into real time space at which point the kernel API, userland C runtime and all sorts have to change.
Plus there are good reasons not to arbitrarily sleep for sub-second times for the sake of the scheduler. You probably should be using select or other event driven programming models rather than relying on delays.
Fwiw, on recent FreeBSDs nanosleep(2) is the system call, which is what both sleep(3) and sleep(1) use, the former with whole numbers of seconds, and the latter with fractional seconds.
15 comments
[ 2.6 ms ] story [ 45.1 ms ] threadhttp://www.quora.com/Hacker-Culture/Who-was-Brett-Glass-as-n...
Brett Glass himself shows up to answer.
http://www.cringely.com/tag/brett-glass/
http://lists.freebsd.org/pipermail/freebsd-chat/2003-March/0...
I read Poul-Henning's argument, which did not say that ideas should be dismissed, but rather 'just built', as attempting to contribute to a constructive spirit, and Glass's argument as defending the broken management process.
Plus there are good reasons not to arbitrarily sleep for sub-second times for the sake of the scheduler. You probably should be using select or other event driven programming models rather than relying on delays.
http://bikeshed.org/
(The same site also seems to be on bikeshed.com, which is where I remembered it)